B.Tech AI & Data Science graduate, RAIT, D.Y. Patil University, Class of 2026. I care about the whole loop, from a model's confusion matrix to the button a person actually presses.
PROJECT |
WHAT IT IS |
BUILT WITH |
|---|---|---|
| AMINAL | Local-first voice assistant for Windows. Voice, vision and memory on your own GPU, nothing leaves the machine | Python Whisper Ollama |
| Fill.ai | AI form filler for Chrome. Learns you from your resume, fills what it knows, asks about the rest, never submits | JavaScript Chrome MV3 LLMs |
| PixelForge | Leukemia detection on C-NMC at 0.90 ROC-AUC, a custom GAN augments scarce medical data | PyTorch GANs |
| MindFirst | Mental-health screening for 100+ students, chatbot plus psychometrics at 95% accuracy | Next.js Flask Prisma |
| Rail Mitra | Railway platform and track assignment, solved as constraint programming | Python OR-Tools |
| IntelliStatement | Bank statements in, financial insight out. LLM extraction from PDFs and images | FastAPI Streamlit LLM |
| Sim Sim | Voice-driven desktop assistant with a native GUI, 70% faster task execution | Python PySide6 |
| Committee Connect | Task and attendance tracking for college committees, real-time sync | Android Java Firebase |
Running in a browser right now: PixelShift, batch image conversion with zero uploads, and To-Do, full-stack task boards with auth. The rest, including the site itself, lives at mosambiswas.com.
WHERE |
WHAT |
WHEN |
|---|---|---|
| Sedna Technologies | Software Developer Intern. Work across 4 client projects, plus a full client-facing dashboard built solo | Jan / Apr 2026 |
| Samsung Innovation Campus | AI Trainee. Four-month industry-academia program, co-certified by Samsung Southwest Asia & TSSC | Nov 2025 / Mar 2026 |
| Photocircle RAIT | Technical Head. 3 workshops (75+ students), 4 photo walks, 2 inter-college competitions, participation up 60% | 2022 / 2025 |
Comparative Analysis of Psychometric Data for Assessing Mental Health Testing Needs Using Machine Learning Models · IEEE · ICAIIHI 2025.
Can supervised learning decide who should proceed to a full mental-health evaluation? Five models benchmarked on PHQ-9 / GAD-7 responses with cross-validation and SHAP interpretability. Logistic regression came out near-perfect and transparent enough to be clinically meaningful. In a clinic, explainable beats clever.
ML & Data Python PyTorch scikit-learn XGBoost SHAP Pandas
Web & Mobile JavaScript TypeScript Next.js Node.js Flask FastAPI Android (Java)
Tools Git MySQL Streamlit Figma Photoshop
Currently deep in: LLMs · computer vision · system design
Street, portraits, monochrome, concerts, wildlife. Technical Head, Photocircle RAIT, 2022 / 2025. The archive is SheiChobi, Bengali for “that picture.” 84 photographs, six collections.
© 2026 MOSAM BISWAS · NAVI MUMBAI, IN · ALL RIGHTS RESERVED · NO TEMPLATES, NO WIDGETS · BUILT BY HAND