Interactive Résumé
The Journey of a
Mathematician.
From pure mathematics to cancer genomics, multi-omics, and deployed clinical AI. Each stop on the left rail is a station — scroll, and the train arrives while the story unfolds one frame at a time.
2018 — 2021 · The Foundation
B.Sc. Mathematics
St. Xavier's College, Mumbai · CGPA 9.31 / 10
A rigorous major in Mathematics with a minor in Physics — Linear Algebra, Vector Analysis, Optimization, and Topology. The analytical bedrock behind every model I build, and the reason I think about biology in terms of geometry, structure, and completion.
2020 · Ahead of the Curve
Building the ML Foundation
As early as 2020 — while still an undergraduate — advanced, verified specializations in deep learning and its mathematics.
TensorFlow Developer
Building, training & deploying scalable neural networks for vision and sequence data.
ML on Google Cloud
Production machine-learning pipelines and scalable cloud training end-to-end.
Mathematics for ML
Linear Algebra, Multivariate Calculus, and PCA — the maths under the models.
2021 · First Author
First Research Papers
First independent research — bridging optimization theory and cancer biology as an undergraduate.
Research Paper on Gradient Descent
An analytical study of gradient-based optimization — the mathematical engine behind modern deep learning.
Cancer Cell Mapping — Single-cell RNA-seq
A deep-learning model mapping cancer cells to cancer cell lines and cancer types from single-cell RNA-seq pan-cancer data.
Sep 2021 — Mar 2025 · 3.5 Years
Bioinformatics Research Engineer
BRIC — NIBMG, Kalyani · National Supercomputing Mission
- National Supercomputing Mission (NSM): large-scale genomics, transcriptome mapping, and drug-discovery pipelines on HPC clusters.
- GPU pipeline engineering: scalable, high-performance GPU pipelines for tissue feature extraction from foundation models.
- Histopathology vision: deep convolutional autoencoders & Multiple Instance Learning on whole-slide images.
Ensemble ML evaluation of in-silico pathogenicity predictors for cancer driver mutations
Random-forest ensemble reaching AUC-ROC 0.89 across 502 HNSC patients & 299 driver genes. 10.1186/s13040-024-00420-x
Mar 2023 — Feb 2025 · Germany
M.Sc. Artificial Intelligence
IU International University of Applied Sciences · Grade 1.8
A formal deep-learning education layered on years of applied research — advanced ML, deep learning, and biomedical applications — pursued alongside my full-time engineering role, culminating in a Grade 1.0 thesis (next stop).
June 2023 · Helmholtz AI Conference, Germany
Graph Attention for Multi-Omics
Patient-specific heterogeneous graphs combining genomics, transcriptomics & methylomics — distilled with Graph Attention Autoencoders.
- Heterogeneous multi-omic networks per patient (TCGA / HNSCC).
- Graph Attention Autoencoders in PyTorch Geometric for biomarker discovery.
- First-author research — presented as a flash talk & poster.
Aug 2024 · Under Journal Review
Immune-Omics × Pathology Vision
Predicting high-risk HNSCC directly from whole-slide images — AUC > 0.85 — by aligning pathology features with immune-omics.
- Contrastive feature alignment of pathology visuals ↔ immune-omics via cosine-similarity loss.
- Attention-based MIL head aggregating tile features per WSI (100k × 80k px).
- Clinical impact: high-risk stratification from an H&E slide alone — cutting NGS cost & turnaround.
M.Sc. Thesis · Grade 1.0
RAG-LLMs for Metabolomics
The top-grade thesis became a first-author paper — MedDiscover, published in CSBJ.
MedDiscover: A Domain-Specific RAG Framework for Evidence-Grounded Knowledge Extraction in Metabolomics
Vatsal Pravinbhai Patel, Elena Jolkver, Anne Schwerk · Published 9 April 2026. Thesis graded 1.0 (top) at IU / Berlin Institute of Health.
A domain-specific biomedical Retrieval-Augmented Generation (RAG) method — MedCPT embeddings over metabolic-disorder literature (ICD-10 E70–E88), with generation strictly grounded in retrieved context to curb hallucination, benchmarked on a two-tier Gold + Silver QA suite (RAGAS metrics).
HawkFranklin Research
Founder & Principal Research Engineer · May 2025 — Present
PelliScope · AI Teledermatology
AI for pre-medical dermatology inquiry and rapid infectious screening — validated at AUC ≈ 0.863, developed for Emirates Health Services (EHS) in partnership with Oracle Health.
HawkFranklin's dermatology solution was shortlisted by Emirates Health Services for showcasing at WHX Health Dubai (Feb 9–12, 2026).
For people at home
A calmer way to act on a concern instead of waiting, guessing, or abandoning the process halfway through.
For Hospitals & Intake Teams
Reduces the repetitive back-and-forth that usually happens before a case is ready to review.
For Sovereign Clinics
Cases arrive with better structure, a clearer sense of urgency, and a smoother path into the review queue.
PelliScope EHS Kiosk in Action
The self-service teledermatology kiosk built for Emirates Health Services with Oracle Health — the full patient-to-doctor screening journey, shown live at WHX Health Dubai.
Why it helps
Efficient Tele-Medicine
Workflow comparison
AI triage saves 3 mins of discovery time per patient.
8 patients / hourModel performance (AUC)
Zero-shot comparison on dermatology benchmarks
Unit Economics
Daily compute spend (6 hours)
CPU edge inference
"A single machine costing less than a cup of coffee per day drives this entire workflow."
Daily additional revenue
Commercial Models
Choose a light-touch API path or a personal clinic setup, depending on how much of the experience you want to own.
API call based licensing
Built for self-screening inquiries at scale and headless embedding inside existing patient flows.
Clinic personal setup
Per instance, with the experience open to re-branding.
- Telemedicine-style UI included
- White-label for private networks
- Flexible clinic-facing rollout
Recognised by Emirates Health Services
An official acknowledgment from EHS — the UAE's national health authority — for the PelliScope dermatology solution ahead of the WHX Dubai showcase.
HawkFranklin Research
Flagship Pathology Workstation · Precision Pathology Co-Pilot
01 · The Product
Multimodal AI for End-to-End Oncology
Core Pillars
-
Diagnosis
Molecular Subtype Prediction (Breast, Lung, Oral)
-
Prognosis
Mutational Biomarker Risk & Survival Analysis
-
Treatment Plan
Chemotoxicity Probability & Safety Assessment
Patient #TCGA-BR-8829
Invasive Ductal Carcinoma
Survival Risk
High
0.87 (P<0.001)
Key Mutation
TP53+
Confidence 94%
Massive Scale Training
Trained on TCGA (2.5PB) & validated on CPTAC.
3 Major Cancer Types
Specialized for Breast, Lung, and Oral cancers.
WSI & Molecular Insights
Predicts subtypes & mutations from standard H&E slides.
Pathology Workstation in Action
OncoGemma integrates high-resolution Whole-Slide Imaging with multimodal AI for cancer subtype prediction and mutational-biomarker identification — paired with my Graph Attention Autoencoder for risk stratification. Explore OncoGemma
02 · Resource Optimization
Workflow Acceleration
Save 2–3 Days / PatientPredicts biomarkers directly from standard tissue slides.
Savings: $100 – $1000 / case$10/hr
GPUHigh-Speed Inference
30s
Per Whole-Slide Image (16GB GPU)
"Democratizing precision oncology by removing the barrier of expensive genomic sequencing infrastructure."
04 · Proprietary Technology
WSI Input
Gigapixel Slide
Patch Encoding
Tiling & Features
Graph Attention Autoencoder
Learns spatial cell & morphology relationships
Multi-Output
Subtype · Risk · Mutation
Data Foundation
Rigorously trained on the gold-standard TCGA dataset and validated against CPTAC to ensure generalizability across diverse cohorts.
Hypothetical Performance
High accuracy identifying aggressive subtypes across Breast & Lung cohorts.
05 · OncoGemma Team
Vatsal Patel
Founder & Research Engineer
M.Sc Artificial Intelligence (IU Berlin)
Dr. Abhijeet Patel
Clinical Research Physician
Bachelor of Medicine & Surgery
Dr. Yash Patel
Clinical Research Physician
Bachelor of Medicine & Surgery
Dr. Nishi Seth
Consultant Dermatologist
M.D Dermatology
Saurav Roy
Operations Head
M.Tech Biotechnology
Ananya Pal
Finance & Accounting Head
B.S Economics
Research Manuscript Available on Request Provisional Patent Filed
06 · Vision & Portfolio
Operational Revenue Model
Path A: HFR-Initiated Products
We identify a market gap, build a validated product (like OncoGemma), then seek commercial partners to scale it.
Path B: Partner-Initiated R&D
You bring a defined challenge. We build the dedicated AI solution — acting as your external deep-tech R&D team.
Our Philosophy
HawkFranklin's long-term vision is to innovate science with commercial rigor. Our team spans physicists, chemists, biologists, and AI engineers building breakthroughs across quantum computing, finance, and healthcare.
"We are a resilient, self-funded team. We will succeed with or without external funding — this is an opportunity to join a team that is already moving forward."
Broader Biomedical Portfolio
OncoGemma
AI agents for digital pathology analysis
C-Risq
Multi-omics for cancer risk stratification
MedDiscover
RAG copilot for biomedical Q&A
Clinical FM
Foundation models for tabular EHR data
The Record
Publications
MedDiscover: A Domain-Specific RAG Framework for Evidence-Grounded Knowledge Extraction in Metabolomics
10.34133/csbj.0018An ensemble machine learning-based performance evaluation identifies top in-silico pathogenicity prediction methods for cancer driver mutations
10.1186/s13040-024-00420-xMulti-faceted dysregulated immune response for COVID-19 infection explaining clinical heterogeneity
10.1016/j.cyto.2023.156434Deep Learning model to map cancer cells to cancer cell lines and cancer types from single-cell RNA-seq pan-cancer data
Verified Peer Reviewer
Patents
System and Methods for Secondary Multiomic Profile Synthesis
Systems and Methods for an Adaptive AI Agent-Driven Interactive Learning Ecosystem
Technical Toolchain
References
Dr. Anne Schwerk
Professor of AI, IU International University of Applied Sciences, Bad Honnef, Germany
anne.schwerk@iu.orgProf. Meenal Kolkar
Associate Professor, Dept. of Mathematics, St. Xavier's College, Mumbai, India
meenal.kolkar@xaviers.eduLet's build something that matters.
Clinical partners · investors · researchers on the math of biology.