~/Biswajit-sow — zsh
❯ cat about.mdI build AI systems the way engineers design infrastructure — failure-resilient first, elegant second.
Final year B.Tech CS student at UEM Kolkata (GPA: 8.46), with production-grade AI deployments already in the field. I gravitate toward problems with real consequences — medical diagnostics, road safety intelligence, systems that run at scale for people who depend on them.
Security-minded from day one: every model is a potential attack surface, every data pipeline a liability. Pursuing Google Cybersecurity alongside deep AI work — because understanding how systems fail is as valuable as knowing how to build them.
› PyTorch · LangChain · RAG · MLOps · XGBoost · LangGraph · Agentic AI
› Flask · React · MERN · Streamlit · AWS · GCP · Databricks · Pinecone · Astra DB
| 📦 Repos | 🎓 GPA | 🗃️ Records Processed | 🎯 Model Accuracy |
|---|---|---|---|
| 33 | 8.46 / 10 | 300,000+ | 86% (XGBoost) |
// SKILLS & STACK — proficiency overview
› AI & Machine Learning
| Skill | Proficiency |
|---|---|
Python / SQL / C |
████████████████████ 90% |
LangChain / RAG / Agentic AI |
█████████████████░░░ 85% |
PyTorch / Scikit-Learn |
████████████████░░░░ 82% |
XGBoost / ML Pipelines |
████████████████░░░░ 82% |
MLOps / LangSmith |
███████████████░░░░░ 75% |
› Frontend & Deployment
| Skill | Proficiency |
|---|---|
Flask / Streamlit |
████████████████░░░░ 80% |
React / MERN Stack |
██████████████░░░░░░ 72% |
Gradio |
██████████████░░░░░░ 70% |
Next.js / Node.js |
█████████████░░░░░░░ 65% |
› Data Visualization & Analytics
| Skill | Proficiency |
|---|---|
Matplotlib / Seaborn / NumPy |
█████████████████░░░ 85% |
Power BI / Dashboards |
███████████████░░░░░ 78% |
Plotly / Interactive Viz |
███████████████░░░░░ 75% |
Grad-CAM / XAI Visualization |
██████████████░░░░░░ 72% |
Databricks / Spark |
█████████████░░░░░░░ 68% |
› Databases & Vector Stores
| Skill | Proficiency |
|---|---|
SQL / SQLite |
████████████████░░░░ 82% |
Pinecone (Vector DB) |
███████████████░░░░░ 78% |
Astra DB (Cassandra) |
███████████████░░░░░ 75% |
MongoDB |
██████████████░░░░░░ 70% |
Neo4j / Graph DB |
████████████░░░░░░░░ 60% |
› Cloud & Infrastructure
| Skill | Proficiency |
|---|---|
AWS / GCP |
█████████████░░░░░░░ 68% |
Docker / CI-CD |
████████████░░░░░░░░ 60% |
|
XGBoost accident severity prediction (86% acc) on 300K+ records. Geospatial hotspot engine with DBSCAN + an AI Copilot via LangGraph for real-time safe route recommendations. Full-stack React-Flask app with 25+ risk features. |
CNN pneumonia detection from chest X-rays with Grad-CAM saliency maps. Llama-3 clinical reasoning engine. Real-time explainable diagnostics in ~2 seconds — heatmaps + confidence scores + clinical report. |
|
Full-stack MERN app covering 9+ mental health areas and 30+ technical domains — powered by expert-specific prompt-engineered LLMs. Strict domain separation across 39 knowledge areas. Reduced hallucinations by 90% and improved accuracy by 85% via flowchart-driven response routing. |
|
- ▸ 📄 Co-authored "Harnessing the True Potential of LLMs" — published in Springer
- ▸ 🏆 Top 60 of 8,000 participants — AICTE & Edunet Foundation AI & Data Science Program (April 2025)
- ▸ ⚡ Led backend in 48-hour sprint at Hack4Bengal S4 — 4-member team, full AI solution delivered
| Certification | Issuer | Badge |
|---|---|---|
| Machine Learning Models in Microsoft Azure | Microsoft |  |
| Google Cybersecurity |  |
