M.Tech, Structural & Computational Biology — IIT Roorkee
ex-Project Associate, IISc Bangalore
I build deep learning systems and AI agents for medical imaging and clinical data —
pipelines that work on real clinical problems, not just benchmark datasets.
- Medical Image Analysis — H&E histopathology, WSI segmentation, Attention UNet, active learning pipelines
- MLOps & Deployment — FastAPI · Celery · Redis · Docker · AWS EC2 · Streamlit
- LLMOps & AI Agents — LangGraph · ChromaDB · RAG · LangChain · Groq API · LLaMA 3
- Data Science — Python, Pandas, SQL, statistical analysis, EDA
- Bioinformatics — RNA-seq pipelines, DESeq2, genomic data processing
| Project | Description | Stack |
|---|---|---|
| 🔒 Cardiac Histopathology Segmentation | Attention UNet for inflammation detection in 6,600+ cardiac WSIs — 77.9% defensible Dice (data-leakage corrected), cascaded dual-threshold inference, FastAPI + Celery + Redis deployment (confidential thesis) | PyTorch · Attention UNet · OpenCV · FastAPI · Celery · Redis · Docker |
| Medical Knowledge Agent | LangGraph autonomous AI agent routing natural language queries to SQLite patient database or ChromaDB medical PDF vector store — Next.js frontend, FastAPI backend, Docker Compose on AWS EC2 | LangGraph · ChromaDB · SQLite · FastAPI · Next.js · Docker · Groq API |
| Brain Tumor MLOps | Full-stack MRI classification deployed live on AWS EC2 (eu-north-1) — PyTorch CNN served via FastAPI, Streamlit frontend, Docker Compose | PyTorch · FastAPI · Streamlit · Docker · AWS EC2 |
| AI Terminal Chat | Terminal-based AI assistant with file-reading capability using Groq API and LLaMA 3 | Python · Groq API · LLaMA 3 |
Deep Learning PyTorch · Attention UNet · CNNs · OpenCV · CUDA · scikit-learn
MLOps FastAPI · Celery · Redis · Docker · Docker Compose · AWS EC2 · Streamlit
LLMOps LangGraph · ChromaDB · LangChain · Groq API · LLaMA 3 · RAG
Data Science Python · Pandas · NumPy · SQL · SQLite · R · Bash/Shell
Bioinformatics RNA-seq · DESeq2 · BED/BAM · Genomic Pipelines · ggplot2
Tools Git/GitHub · Linux/Unix · SSH · Jupyter · Google Colab · Arduino
- 🔬 Corrected a train/validation data-leakage flaw in production UNet pipeline — replaced an inflated 89.6% Dice score with a scientifically defensible 77.9% generalisation metric on 6,600+ cardiac WSIs
- 🧬 2 years at IISc Bangalore — contributed to a point-of-care PCR diagnostic prototype for sickle cell anaemia detection that progressed to real clinical evaluation
- 🚀 Deployed multiple live AI systems on AWS EC2 — brain tumour classification and medical knowledge agent, both containerised with Docker Compose
- 💡 Built autonomous LangGraph agent combining SQL patient data queries and RAG medical document retrieval in a single natural language interface
📧 sachinkumar_p@bt.iitr.ac.in
💼 LinkedIn
🌐 Portfolio
🏫 IIT Roorkee | ex-IISc Bangalore
Open to remote freelance projects in medical imaging AI, MLOps, and LLM-powered systems.