PepSys
PepSys
From Dr. Reddy’s Cambridge R&D Site to FDA Submission
From Dr. Reddy’s Cambridge R&D Site
to FDA Submission
PepSys
From Dr. Reddy’s Cambridge R&D Site to FDA Submission
Peptide Purification Reimagined:
Peptide Purification Reimagined:
Peptide Purification Reimagined:
AI and Bioinformatics Solving
a $3B Industry Challenge
AI and Bioinformatics Solving
a $3B Industry Challenge
AI and Bioinformatics Solving
a $3B Industry Challenge
Collaborators:
2 Scientists, 1 designer (me), 1 ML engineer, 2 PMs
Collaborators:
2 Scientists, 1 designer,
1 ML engineer, 2 PMs
Collaborators:
2 Scientists, 1 designer (me),
1 ML engineer, 2 PMs
Form of Application:
PharmaTech,
Web Application
Form of Application:
PharmaTech,
Web Application
Form of Application:
PharmaTech,
Web Application
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
Duration:
1 and half months
Duration:
1 and half months
Duration:
1 and half months
Context
Context
Context
Partnering with Dr. Reddy’s Cambridge R&D team, I contributed to the design and visualisation of an AI-based platform built on peptide synthesis model for impurity prediction and comparison. The model was developed as part of Dr. Reddy’s internal research initiative under the Peptide Synthesis Chemical Method and is currently progressing toward FDA model master file submission.
Partnering with Dr. Reddy’s Cambridge R&D team, I contributed to the design and visualisation of an AI-based platform built on peptide synthesis model for impurity prediction and comparison. The model was developed as part of Dr. Reddy’s internal research initiative under the Peptide Synthesis Chemical Method and is currently progressing toward FDA model master file submission.
My Key Contributions
My Key Contributions
Led cross-functional workshops to align business goals, user needs and technical constraints.
Simplified a complex scientific process into a usable AI based digital workflow.
Designed the UI/UX for model interaction, impurity comparison, and reference validation.
Note: Following the company policy and NDA, Certain details are intentionally altered for privacy. The reflections shared
here come from my personal understanding, reflection and learnings.
Note: Following the company policy and NDA, Certain details are intentionally altered for privacy. The reflections shared here come from my personal understanding, reflection and learnings.
What are Peptides?
What are Peptides?
What are Peptides?
Short chains of amino acids linked by peptide bonds
Short chains of amino acids linked by peptide bonds
Similar to protein but shorter in length (2 to 50 amino acids)
Similar to protein but shorter in length (2 to 50 amino acids)
They act as hormones, neurotransmitters, enzymes
They act as hormones, neurotransmitters,
enzymes
It is used to treat conditions like cancer, obesity, diabetes, hormonal disorders
It is used to treat conditions like cancer,
obesity, diabetes, hormonal disorders
High specificity and minimal side effects
High specificity and minimal side effects
RLD (Reference Listed Drug): The approved reference product used for comparison to ensure safety, quality, and effectiveness.
Problem Context
Problem Context
Problem Context
Meet Dr. Samuel — a passionate scientist who loves his work but is tired of the endless trial and error. No matter how hard he tries, something always slows him down — delays, rework, more paperwork.
How would Pepsys help Dr. Samuel make his life easy?
How would Pepsys help Dr. Samuel make his life easy?
How would Pepsys help Dr. Samuel
make his like easy?
Step – 1: Upload & Identify reference drug-All set for analysis
Step – 1: Upload & Identify reference drug-All set for analysis
Step – 1: Upload & Identify reference drug-All set for analysis
Upload the peptide file. AI detects the Reference Listed Drug (RLD) — review and set for analysis.
Upload the peptide file. AI detects the Reference Listed Drug (RLD) — review and set for analysis.
Step – 2: Analysis & quick Insights
Step – 2: Analysis & quick Insights
Step – 2: Analysis & quick Insights
View comprehensive results in tables and visualizations. Compare with RLD and explore AI-generated insights.
View comprehensive results in tables and visualizations. Compare with RLD
and explore AI-generated insights.
View comprehensive results in tables and visualizations. Compare with RLD and explore AI-generated insights.
Step – 3: Unified Dashboard
Step – 3: Unified Dashboard
Step – 3: Unified Dashboard
All reports stored in one centralized location. Search, filter, and access anytime.
View comprehensive results in tables and visualizations. Compare with RLD
and explore AI-generated insights.
All reports stored in one centralized location. Search, filter, and access anytime.
Value added
Value added
Value added
The platform was recognized at the FDA Center for Research on Complex Generics (CRCG) workshop for measurably streamlining peptide impurity analysis — Now, Dr. Reddy’s is in the process of applying the model master file for regulatory submission
— marking a major step toward AI-driven peptide development.
The platform was recognized at the FDA Center for Research on Complex Generics (CRCG) workshop for measurably streamlining peptide impurity analysis — Now, Dr. Reddy’s is in the process of applying the model master file for regulatory submission — marking a major step toward AI-driven peptide development.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Let’s go
Let’s go
Let’s get a simple understanding of peptide impurities first
Let’s get a simple understanding of peptide impurities first
Let’s get a simple understanding
of peptide impurities first
What peptide synthesis means
Peptide synthesis is the process of combining amino acids to create small protein-like molecules used in advanced medicines.
What peptide synthesis means
Peptide synthesis is the process of combining amino acids to create small protein-like molecules used in advanced medicines.
How and why impurities are formed
During synthesis, unwanted by-products — called impurities — form because of chemical reactions, equipment conditions, or peptide sequence complexity.
How and why impurities are formed
During synthesis, unwanted by-products — called impurities — form because of chemical reactions, equipment conditions, or peptide sequence complexity.
Why is impurity detection in peptide synthesis critically important?
Why is impurity detection in peptide synthesis critically important?
Users POV 01
Impurities in peptides can harm patients, causing toxicity or reduced drug effectiveness.
01 Users POV
Impurities in peptides can harm patients, causing toxicity
or reduced drug effectiveness.
Users POV 01
Impurities in peptides can harm patients, causing toxicity or reduced drug effectiveness.
Regulatory POV 02
Drugs with unacceptable impurity levels can face rejection, delayed approval, or mandatory reformulation.
02 Regulatory POV
Drugs with unacceptable impurity levels can face rejection, delayed approval,
or mandatory reformulation.
Regulatory POV 02
Drugs with unacceptable impurity levels can face rejection, delayed approval, or mandatory reformulation.
Business POV 03
Impurity issues can cost companies millions and risk a $3B market opportunity.
03 Business POV
Impurity issues can cost companies millions and risk
a $3B market opportunity.
Business POV 03
Impurity issues can cost companies millions and risk a $3B market opportunity.
Manual impurity detection is exhausting and time-consuming.
What if AI could handle the heavy lifting for us?
Manual impurity detection is exhausting and time-consuming.
What if AI could handle the heavy lifting for us?
Manual impurity detection is exhausting
and time-consuming.
What if AI could handle the heavy
lifting for us?
Initial flow proposed by business
Initial flow proposed by business
Initial flow proposed by business
What we received wasn’t enough
What we received wasn’t enough
Too complex. Too much jargon.
No clarity on the actual user journey.
AI’s role is missing from the workflow.
Too complex.
Too much jargon.
No clarity on the
actual user journey.
AI’s role is missing
from the workflow.
Understanding the core problem through a design workshop
Understanding the core problem through a design workshop
Understanding the core problem
through a design workshop
Agenda
Agenda
To align the team, build clarity, and uncover the true design opportunities, we conducted a focused design workshop with a team of scientists and product manager. The workshop was structured to bring all stakeholders—scientists, analysts, and product teams—onto the same page and to simplify the complexity surrounding peptide impurity prediction.
To align the team, build clarity, and uncover the true design opportunities, we conducted a focused design workshop with a team of scientists and product manager. The workshop was structured to bring all stakeholders—scientists, analysts, and product teams—onto the same page and to simplify the complexity surrounding peptide impurity prediction.
Objectives
Objectives
01
01
To simplify
the jargons
To simplify the jargons
02
To understand the
process flow
03
To understand the user
base and their behaviour
and pain points
04
To understand how
it is integrated in organisational system
02
To understand the
process flow
03
To understand the user base
and their behaviour and pain points
04
To understand how it is
integrated in organisational system
6
6
Participants
Participants
60
60
Minutes
Minutes
250+
250+
Post its
Post its
De-cluttering the chaos
De-cluttering the chaos
De-cluttering the chaos
We understood an overview of the entire peptide value chain and how early impurity detection helps
achieve optimisation.
We understood an overview of the entire peptide value chain and how early impurity detection helps achieve optimisation.
We understood an overview of the entire peptide
value chain and how early impurity detection helps
achieve optimisation.
“Peptide therapeutics have huge market value for DRL. Peptides can provide revenue of up to $3Bn.”
~ Product manager
“Peptide therapeutics have huge market value for DRL. Peptides can provide revenue of
up to $3Bn.” ~ Product manager
“Peptide therapeutics have huge market value
for DRL. Peptides can provide revenue of up
to $3Bn.” ~ Product manager
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
We discovered that the challenges in peptide development were not isolated issues – they were part of interconnected loops
that kept reinforcing delays, high costs, and inefficiencies.
We discovered that the challenges in peptide development were not isolated issues – they were part of interconnected loops
that kept reinforcing delays, high costs, and inefficiencies.
Legends
C
Critical Impact Factors
Critical Impact Factors
T
Technical Inefficiencies
Technical Inefficiencies
O
Operational Issues
Operational Issues
+ causal relationship
+ causal relationship
R1: Technology-Delay
R1: Technology-Delay
R2: Quality-Cost
R2: Quality-Cost
R3: Market-Capacity
R3: Market-Capacity
Key Reinforcing loops identified
Key Reinforcing loops identified
Key Reinforcing loops identified
01
01
R1: Technology–Delay Loop
R1: Technology–Delay Loop
Inefficient technologies
Inefficient technologies
Manual trial & error
Manual trial & error
Resources underutilized
Resources underutilized
More delays
More delays
Development timeline stretches
Development timeline stretches
Late Filing
Late Filing
02
R2: Quality–Cost Loop
R2: Quality–Cost Loop
Poor quality & more impurities
Poor quality & more impurities
Batch failures
Batch failures
More rework
More rework
Higher COGS
Higher COGS
Compromised quality
Compromised quality
More cost issues
More cost issues
03
03
R3: Market–Capacity Loop
R3: Market–Capacity Loop
Underutilized resources
Underutilized resources
Inefficient value chain
Inefficient value chain
More deficiencies
More deficiencies
Approval delays
Approval delays
Can’t meet market needs
Can’t meet market needs
FTM Compromised
FTM Compromised
Synthesising what all loops together reveal
Synthesising what all loops together reveal
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
How can we make the existing system efficient?
How can we make the existing system efficient?
How can we make the existing
system efficient?
We studied the current workflow to pinpoint redundant steps and opportunities to make the process faster.
We studied the current workflow to pinpoint redundant steps and opportunities
to make the process faster.
We studied the current workflow to pinpoint redundant steps and opportunities to make the process faster.
Dr. Samuel is representing our primary persons:
Dr. Samuel is representing our primary persons:
API Scientists
Formulation scientists
Scale up scientists
Formulation manufacturing team
R&D team
Biologics team
Fermentation scientists
Clinical scientists
Let’s look at Dr. Samuel’s existing journey:
Let’s look at Dr. Samuel’s existing journey:
To-be user flow (Human and AI Collaboration)
To-be user flow (Human and AI Collaboration)
To-be user flow (Human and
AI Collaboration)
Ideation
Ideation
Ideation
During ideation, the core workflow itself was intentionally simple: upload the peptide file, trigger analysis, and review results. The real complexity emerged in navigation.
During ideation, the core workflow itself was intentionally simple: upload the peptide file,
trigger analysis, and review results. The real complexity emerged in navigation.
During ideation, the core workflow itself was intentionally simple: upload the peptide file, trigger analysis, and review results. The real complexity emerged in navigation.
HMW statement to structure our direction
HMW statement to structure our direction
How might we design a navigation experience that keeps scientists oriented and confident while starting a new
peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
How might we design a navigation experience that keeps scientists oriented and confident while starting a new peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
How might we design a navigation experience that keeps scientists oriented and confident while starting a new peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
Initial explorations
Initial explorations
Idea-1 : Split view
Idea-2 : Separate tab view
Idea-3 : Navigating through side panel
Idea-1 : Split view
Idea-1 : Split view
Idea-2 : Separate tab view
Idea-2 : Separate tab view
Idea-3 : Navigating through side panel
Idea-3 : Navigating through side panel
The early explorations did not work out:
The early ideas looked good visually, but they didn’t match how scientists actually work. They either split attention, added extra clicks, or focused too much on creating new work instead of continuing existing projects. What users really needed was quick access to past work, clear structure, and less switching between screens.
The early explorations did not work out:
The early ideas looked good visually, but they didn’t match how scientists actually work. They either split attention, added extra clicks, or focused too much on creating new work instead of continuing existing projects. What users really needed was quick access to past work, clear structure, and less switching between screens.
Final Option
Final Option
Idea-4 : All-in-all dashboard view
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard
let them quickly scan, check status,
and continue existing projects without
extra steps or confusion.
Idea-4 : All-in-all dashboard view
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard let
them quickly scan, check status, and
continue existing projects without extra
steps or confusion.
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard
let them quickly scan, check status,
and continue existing projects without
extra steps or confusion.
Ideation to final solution
Ideation to final solution
Ideation to final solution
Personalized Onboarding for First-Time Users
Personalized Onboarding for First-Time Users
Personalized Onboarding for
First-Time Users
I introduced a personalized onboarding experience to make first-time users feel guided and confident from
the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring
they have the correct format saved locally before beginning.
I introduced a personalized onboarding experience to make first-time users feel guided and confident from the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring they have the correct format saved locally before beginning.
I introduced a personalized onboarding experience to make first-time users feel guided and confident from the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring they have the correct format saved locally before beginning.
Proactive Error Validation After Upload
Proactive Error Validation After Upload
Proactive Error Validation
After Upload
After upload, the system validates the file and shows a clear summary of detected sequences and impurities.
Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the
data before proceeding.
After upload, the system validates the file and shows a clear summary of detected sequences and impurities. Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the data before proceeding.
After upload, the system validates the file and shows a clear summary of detectedsequences and impurities. Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the data before proceeding.
Manual Verification of AI-Identified RLD
Manual Verification of AI-Identified RLD
Manual Verification of
AI-Identified RLD
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before
proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit
the selection if needed before starting the analysis.
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit the selection
if needed before starting the analysis.
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit the selection if needed before starting the analysis.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm before proceeding.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm
before proceeding.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm before proceeding.
Once confirmed, the system validates the reference drug (RLD) and enables the user to start the analysis.
Once confirmed, the system validates the reference drug (RLD) and enables the user
to start the analysis.
Once confirmed, the system validates the reference drug (RLD) and enables the user to start the analysis.
Progressive Analysis Loader for Better Engagement
Progressive Analysis Loader for Better Engagement
Progressive Analysis Loader for Better Engagement
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into
visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system
processes data in the background.
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system
processes data in the background.
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system processes data in the background.
Comprehensive, Multi-View Analysis Output
Comprehensive, Multi-View Analysis Output
Comprehensive, Multi-View
Analysis Output
The output is presented in two powerful formats — a detailed comparison table and an interactive 3D visualization.
Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review.
The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper
analysis without switching contexts.
The output is presented in two powerful formats — a detailed comparison table and an interactive 3D visualization. Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review. The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper analysis without switching contexts.
The output is presented in two powerful formats —
a detailed comparison table and an interactive 3D visualization. Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review. The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper analysis without switching contexts.
AI-Generated Insights & Recommendations
AI-Generated Insights & Recommendations
AI-Generated Insights
& Recommendations
Alongside the comparative analysis, AI provides structured inferences throughexpandable sections — highlights
structural and property impacts, and suggests recommended methods — helping users understand discrepancies
quickly and take informed action.
Alongside the comparative analysis, AI provides structured inferences through expandable sections — highlights structural and property impacts, and suggests recommended methods — helping users understand discrepancies quickly and take informed action.
Alongside the comparative analysis, AI provides structured inferences through expandable sections — highlights structural and property impacts, and suggests recommended methods — helping users understand discrepancies quickly and take informed action.
Reflection & Closing note
Reflection & Closing note
Reflection & Closing note
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this
work feel very real and meaningful to me.
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this work feel very real and meaningful to me.
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this
work feel very real and meaningful to me.
PepSys
PepSys
From Dr. Reddy’s Cambridge R&D Site to FDA Submission
From Dr. Reddy’s Cambridge R&D Site
to FDA Submission
PepSys
From Dr. Reddy’s Cambridge R&D Site to FDA Submission
Peptide Purification Reimagined:
Peptide Purification Reimagined:
Peptide Purification Reimagined:
AI and Bioinformatics Solving
a $3B Industry Challenge
AI and Bioinformatics Solving
a $3B Industry Challenge
AI and Bioinformatics Solving
a $3B Industry Challenge
Collaborators:
2 Scientists, 1 designer (me), 1 ML engineer, 2 PMs
Collaborators:
2 Scientists, 1 designer,
1 ML engineer, 2 PMs
Collaborators:
2 Scientists, 1 designer (me),
1 ML engineer, 2 PMs
Form of Application:
PharmaTech,
Web Application
Form of Application:
PharmaTech,
Web Application
Form of Application:
PharmaTech,
Web Application
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
My Role:
Research, Workshop,
System Design,
UI design, AI Experience
Duration:
1 and half months
Duration:
1 and half months
Duration:
1 and half months
Context
Context
Context
Partnering with Dr. Reddy’s Cambridge R&D team, I contributed to the design and visualisation of an AI-based platform built on peptide synthesis model for impurity prediction and comparison. The model was developed as part of Dr. Reddy’s internal research initiative under the Peptide Synthesis Chemical Method and is currently progressing toward FDA model master file submission.
Partnering with Dr. Reddy’s Cambridge R&D team, I contributed to the design and visualisation of an AI-based platform built on peptide synthesis model for impurity prediction and comparison. The model was developed as part of Dr. Reddy’s internal research initiative under the Peptide Synthesis Chemical Method and is currently progressing toward FDA model master file submission.
My Key Contributions
My Key Contributions
Led cross-functional workshops to align business goals, user needs and technical constraints.
Simplified a complex scientific process into a usable AI based digital workflow.
Designed the UI/UX for model interaction, impurity comparison, and reference validation.
Note: Following the company policy and NDA, Certain details are intentionally altered for privacy. The reflections shared
here come from my personal understanding, reflection and learnings.
Note: Following the company policy and NDA, Certain details are intentionally altered for privacy. The reflections shared here come from my personal understanding, reflection and learnings.
What are Peptides?
What are Peptides?
What are Peptides?
Short chains of amino acids linked by peptide bonds
Short chains of amino acids linked by peptide bonds
Similar to protein but shorter in length (2 to 50 amino acids)
Similar to protein but shorter in length (2 to 50 amino acids)
They act as hormones, neurotransmitters, enzymes
They act as hormones, neurotransmitters,
enzymes
It is used to treat conditions like cancer, obesity, diabetes, hormonal disorders
It is used to treat conditions like cancer,
obesity, diabetes, hormonal disorders
High specificity and minimal side effects
High specificity and minimal side effects
RLD (Reference Listed Drug): The approved reference product used for comparison to ensure safety, quality, and effectiveness.
Problem Context
Problem Context
Problem Context
Meet Dr. Samuel — a passionate scientist who loves his work but is tired of the endless trial and error. No matter how hard he tries, something always slows him down — delays, rework, more paperwork.
How would Pepsys help Dr. Samuel make his life easy?
How would Pepsys help Dr. Samuel make his life easy?
How would Pepsys help Dr. Samuel
make his like easy?
Step – 1: Upload & Identify reference drug-All set for analysis
Step – 1: Upload & Identify reference drug-All set for analysis
Step – 1: Upload & Identify reference drug-All set for analysis
Upload the peptide file. AI detects the Reference Listed Drug (RLD) — review and set for analysis.
Upload the peptide file. AI detects the Reference Listed Drug (RLD) — review and set for analysis.
Step – 2: Analysis & quick Insights
Step – 2: Analysis & quick Insights
Step – 2: Analysis & quick Insights
View comprehensive results in tables and visualizations. Compare with RLD and explore AI-generated insights.
View comprehensive results in tables and visualizations. Compare with RLD
and explore AI-generated insights.
View comprehensive results in tables and visualizations. Compare with RLD and explore AI-generated insights.
Step – 3: Unified Dashboard
Step – 3: Unified Dashboard
Step – 3: Unified Dashboard
All reports stored in one centralized location. Search, filter, and access anytime.
View comprehensive results in tables and visualizations. Compare with RLD
and explore AI-generated insights.
All reports stored in one centralized location. Search, filter, and access anytime.
Value added
Value added
Value added
The platform was recognized at the FDA Center for Research on Complex Generics (CRCG) workshop for measurably streamlining peptide impurity analysis — Now, Dr. Reddy’s is in the process of applying the model master file for regulatory submission
— marking a major step toward AI-driven peptide development.
The platform was recognized at the FDA Center for Research on Complex Generics (CRCG) workshop for measurably streamlining peptide impurity analysis — Now, Dr. Reddy’s is in the process of applying the model master file for regulatory submission — marking a major step toward AI-driven peptide development.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Above was a high-level project overview; next is a deep dive into problem approach, research insights, and design iterations.
Let’s go
Let’s go
Let’s get a simple understanding of peptide impurities first
Let’s get a simple understanding of peptide impurities first
Let’s get a simple understanding
of peptide impurities first
What peptide synthesis means
Peptide synthesis is the process of combining amino acids to create small protein-like molecules used in advanced medicines.
What peptide synthesis means
Peptide synthesis is the process of combining amino acids to create small protein-like molecules used in advanced medicines.
How and why impurities are formed
During synthesis, unwanted by-products — called impurities — form because of chemical reactions, equipment conditions, or peptide sequence complexity.
How and why impurities are formed
During synthesis, unwanted by-products — called impurities — form because of chemical reactions, equipment conditions, or peptide sequence complexity.
Why is impurity detection in peptide synthesis critically important?
Why is impurity detection in peptide synthesis critically important?
Users POV 01
Impurities in peptides can harm patients, causing toxicity or reduced drug effectiveness.
01 Users POV
Impurities in peptides can harm patients, causing toxicity
or reduced drug effectiveness.
Users POV 01
Impurities in peptides can harm patients, causing toxicity or reduced drug effectiveness.
Regulatory POV 02
Drugs with unacceptable impurity levels can face rejection, delayed approval, or mandatory reformulation.
02 Regulatory POV
Drugs with unacceptable impurity levels can face rejection, delayed approval,
or mandatory reformulation.
Regulatory POV 02
Drugs with unacceptable impurity levels can face rejection, delayed approval, or mandatory reformulation.
Business POV 03
Impurity issues can cost companies millions and risk a $3B market opportunity.
03 Business POV
Impurity issues can cost companies millions and risk
a $3B market opportunity.
Business POV 03
Impurity issues can cost companies millions and risk a $3B market opportunity.
Manual impurity detection is exhausting and time-consuming.
What if AI could handle the heavy lifting for us?
Manual impurity detection is exhausting and time-consuming.
What if AI could handle the heavy lifting for us?
Manual impurity detection is exhausting
and time-consuming.
What if AI could handle the heavy
lifting for us?
Initial flow proposed by business
Initial flow proposed by business
Initial flow proposed by business
What we received wasn’t enough
What we received wasn’t enough
Too complex. Too much jargon.
No clarity on the actual user journey.
AI’s role is missing from the workflow.
Too complex.
Too much jargon.
No clarity on the
actual user journey.
AI’s role is missing
from the workflow.
Understanding the core problem through a design workshop
Understanding the core problem through a design workshop
Understanding the core problem
through a design workshop
Agenda
Agenda
To align the team, build clarity, and uncover the true design opportunities, we conducted a focused design workshop with a team of scientists and product manager. The workshop was structured to bring all stakeholders—scientists, analysts, and product teams—onto the same page and to simplify the complexity surrounding peptide impurity prediction.
To align the team, build clarity, and uncover the true design opportunities, we conducted a focused design workshop with a team of scientists and product manager. The workshop was structured to bring all stakeholders—scientists, analysts, and product teams—onto the same page and to simplify the complexity surrounding peptide impurity prediction.
Objectives
Objectives
01
01
To simplify
the jargons
To simplify the jargons
02
To understand the
process flow
03
To understand the user
base and their behaviour
and pain points
04
To understand how
it is integrated in organisational system
02
To understand the
process flow
03
To understand the user base
and their behaviour and pain points
04
To understand how it is
integrated in organisational system
6
6
Participants
Participants
60
60
Minutes
Minutes
250+
250+
Post its
Post its
De-cluttering the chaos
De-cluttering the chaos
De-cluttering the chaos
We understood an overview of the entire peptide value chain and how early impurity detection helps
achieve optimisation.
We understood an overview of the entire peptide value chain and how early impurity detection helps achieve optimisation.
We understood an overview of the entire peptide
value chain and how early impurity detection helps
achieve optimisation.
“Peptide therapeutics have huge market value for DRL. Peptides can provide revenue of up to $3Bn.”
~ Product manager
“Peptide therapeutics have huge market value for DRL. Peptides can provide revenue of
up to $3Bn.” ~ Product manager
“Peptide therapeutics have huge market value
for DRL. Peptides can provide revenue of up
to $3Bn.” ~ Product manager
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
“Impurities are a bottleneck for getting approval. Avoiding impurities across synthesis and purification is important.”
~ Formulation scientist
We discovered that the challenges in peptide development were not isolated issues – they were part of interconnected loops
that kept reinforcing delays, high costs, and inefficiencies.
We discovered that the challenges in peptide development were not isolated issues – they were part of interconnected loops
that kept reinforcing delays, high costs, and inefficiencies.
Legends
C
Critical Impact Factors
Critical Impact Factors
T
Technical Inefficiencies
Technical Inefficiencies
O
Operational Issues
Operational Issues
+ causal relationship
+ causal relationship
R1: Technology-Delay
R1: Technology-Delay
R2: Quality-Cost
R2: Quality-Cost
R3: Market-Capacity
R3: Market-Capacity
Key Reinforcing loops identified
Key Reinforcing loops identified
Key Reinforcing loops identified
01
01
R1: Technology–Delay Loop
R1: Technology–Delay Loop
Inefficient technologies
Inefficient technologies
Manual trial & error
Manual trial & error
Resources underutilized
Resources underutilized
More delays
More delays
Development timeline stretches
Development timeline stretches
Late Filing
Late Filing
02
R2: Quality–Cost Loop
R2: Quality–Cost Loop
Poor quality & more impurities
Poor quality & more impurities
Batch failures
Batch failures
More rework
More rework
Higher COGS
Higher COGS
Compromised quality
Compromised quality
More cost issues
More cost issues
03
03
R3: Market–Capacity Loop
R3: Market–Capacity Loop
Underutilized resources
Underutilized resources
Inefficient value chain
Inefficient value chain
More deficiencies
More deficiencies
Approval delays
Approval delays
Can’t meet market needs
Can’t meet market needs
FTM Compromised
FTM Compromised
Synthesising what all loops together reveal
Synthesising what all loops together reveal
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
Across all three loops, one pattern stood out clearly: inefficiencies at the early detection stage had a cascading effect on quality, timelines, capacity, and cost.
How can we make the existing system efficient?
How can we make the existing system efficient?
How can we make the existing
system efficient?
We studied the current workflow to pinpoint redundant steps and opportunities to make the process faster.
We studied the current workflow to pinpoint redundant steps and opportunities
to make the process faster.
We studied the current workflow to pinpoint redundant steps and opportunities to make the process faster.
Dr. Samuel is representing our primary persons:
Dr. Samuel is representing our primary persons:
API Scientists
Formulation scientists
Scale up scientists
Formulation manufacturing team
R&D team
Biologics team
Fermentation scientists
Clinical scientists
Let’s look at Dr. Samuel’s existing journey:
Let’s look at Dr. Samuel’s existing journey:
To-be user flow (Human and AI Collaboration)
To-be user flow (Human and AI Collaboration)
To-be user flow (Human and
AI Collaboration)
Ideation
Ideation
Ideation
During ideation, the core workflow itself was intentionally simple: upload the peptide file, trigger analysis, and review results. The real complexity emerged in navigation.
During ideation, the core workflow itself was intentionally simple: upload the peptide file,
trigger analysis, and review results. The real complexity emerged in navigation.
During ideation, the core workflow itself was intentionally simple: upload the peptide file, trigger analysis, and review results. The real complexity emerged in navigation.
HMW statement to structure our direction
HMW statement to structure our direction
How might we design a navigation experience that keeps scientists oriented and confident while starting a new
peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
How might we design a navigation experience that keeps scientists oriented and confident while starting a new peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
How might we design a navigation experience that keeps scientists oriented and confident while starting a new peptide analysis or effortlessly resume past work, without adding cognitive load or unnecessary steps?
Initial explorations
Initial explorations
Idea-1 : Split view
Idea-2 : Separate tab view
Idea-3 : Navigating through side panel
Idea-1 : Split view
Idea-1 : Split view
Idea-2 : Separate tab view
Idea-2 : Separate tab view
Idea-3 : Navigating through side panel
Idea-3 : Navigating through side panel
The early explorations did not work out:
The early ideas looked good visually, but they didn’t match how scientists actually work. They either split attention, added extra clicks, or focused too much on creating new work instead of continuing existing projects. What users really needed was quick access to past work, clear structure, and less switching between screens.
The early explorations did not work out:
The early ideas looked good visually, but they didn’t match how scientists actually work. They either split attention, added extra clicks, or focused too much on creating new work instead of continuing existing projects. What users really needed was quick access to past work, clear structure, and less switching between screens.
Final Option
Final Option
Idea-4 : All-in-all dashboard view
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard
let them quickly scan, check status,
and continue existing projects without
extra steps or confusion.
Idea-4 : All-in-all dashboard view
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard let
them quickly scan, check status, and
continue existing projects without extra
steps or confusion.
Idea 4 worked out:
Idea 4 resonated most because it
felt familiar and aligned with how
users actually work. The dashboard
let them quickly scan, check status,
and continue existing projects without
extra steps or confusion.
Ideation to final solution
Ideation to final solution
Ideation to final solution
Personalized Onboarding for First-Time Users
Personalized Onboarding for First-Time Users
Personalized Onboarding for
First-Time Users
I introduced a personalized onboarding experience to make first-time users feel guided and confident from
the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring
they have the correct format saved locally before beginning.
I introduced a personalized onboarding experience to make first-time users feel guided and confident from the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring they have the correct format saved locally before beginning.
I introduced a personalized onboarding experience to make first-time users feel guided and confident from the start. The flow clearly explains the process and requires users to download the Excel template upfront, ensuring they have the correct format saved locally before beginning.
Proactive Error Validation After Upload
Proactive Error Validation After Upload
Proactive Error Validation
After Upload
After upload, the system validates the file and shows a clear summary of detected sequences and impurities.
Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the
data before proceeding.
After upload, the system validates the file and shows a clear summary of detected sequences and impurities. Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the data before proceeding.
After upload, the system validates the file and shows a clear summary of detectedsequences and impurities. Any errors are explicitly highlighted for correction and re-upload, and users must review and confirm the data before proceeding.
Manual Verification of AI-Identified RLD
Manual Verification of AI-Identified RLD
Manual Verification of
AI-Identified RLD
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before
proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit
the selection if needed before starting the analysis.
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit the selection
if needed before starting the analysis.
After AI detects the likely Reference Listed Drug (RLD), users are required to review and confirm the selection before proceeding. This step ensures scientific accuracy, builds trust in the AI output, and gives users the option to edit the selection if needed before starting the analysis.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm before proceeding.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm
before proceeding.
AI detects the likely reference drug (RLD) and prompts the user to review and confirm before proceeding.
Once confirmed, the system validates the reference drug (RLD) and enables the user to start the analysis.
Once confirmed, the system validates the reference drug (RLD) and enables the user
to start the analysis.
Once confirmed, the system validates the reference drug (RLD) and enables the user to start the analysis.
Progressive Analysis Loader for Better Engagement
Progressive Analysis Loader for Better Engagement
Progressive Analysis Loader for Better Engagement
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into
visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system
processes data in the background.
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system
processes data in the background.
Instead of a generic spinner, I designed a step-by-step progress loader that shows clear analysis stages into visible checkpoints in real time. This keeps users informed, reduces uncertainty, and builds trust while the system processes data in the background.
Comprehensive, Multi-View Analysis Output
Comprehensive, Multi-View Analysis Output
Comprehensive, Multi-View
Analysis Output
The output is presented in two powerful formats — a detailed comparison table and an interactive 3D visualization.
Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review.
The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper
analysis without switching contexts.
The output is presented in two powerful formats — a detailed comparison table and an interactive 3D visualization. Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review. The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper analysis without switching contexts.
The output is presented in two powerful formats —
a detailed comparison table and an interactive 3D visualization. Users can instantly compare results with the RLD, with discrepancies clearly flagged upfront for quick review. The multi-functional screen supports filters, parameter selection, and multiple viewing modes, enabling deeper analysis without switching contexts.
AI-Generated Insights & Recommendations
AI-Generated Insights & Recommendations
AI-Generated Insights
& Recommendations
Alongside the comparative analysis, AI provides structured inferences throughexpandable sections — highlights
structural and property impacts, and suggests recommended methods — helping users understand discrepancies
quickly and take informed action.
Alongside the comparative analysis, AI provides structured inferences through expandable sections — highlights structural and property impacts, and suggests recommended methods — helping users understand discrepancies quickly and take informed action.
Alongside the comparative analysis, AI provides structured inferences through expandable sections — highlights structural and property impacts, and suggests recommended methods — helping users understand discrepancies quickly and take informed action.
Reflection & Closing note
Reflection & Closing note
Reflection & Closing note
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this
work feel very real and meaningful to me.
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this work feel very real and meaningful to me.
This project was definitely a stepping stone for me in understanding how AI intervenes in complex processes. I learned that AI cannot manage everything single-handedly; human intervention remains an essential step for validating AI outcomes.
The solution delivered clear impact, reducing turnaround time by 40% and driving 30% adoption among users.
Presenting the platform at the FDA CRCG workshop and seeing it move toward regulatory submission made the impact of this
work feel very real and meaningful to me.