# Mosam Biswas > Portfolio of Mosam Biswas, a software engineer in Navi Mumbai, India, working across AI / ML and full-stack development. A 2026 B.Tech AI & Data Science graduate of RAIT (D. Y. Patil Deemed to be University) with a software internship, a co-authored IEEE paper, and shipped work that ranges from a fully local voice assistant to an AI form-filling Chrome extension. Open to full-time roles and internships. - Role: Software engineer · AI / ML · Full-stack - Based in: Navi Mumbai, India (IST, UTC+5:30) - Status: Open to full-time roles and internships - Email: mosambiswas999@gmail.com - Everything on this site in one Markdown file: https://www.mosambiswas.com/llms-full.txt ## Profile - [Portfolio](https://www.mosambiswas.com/): the homepage. Twelve selected projects, about, research, photography and a contact form - [Résumé](https://www.mosambiswas.com/resume.html): experience, seven selected projects, research, education, stack, certifications and leadership - [Résumé PDF, one page](https://www.mosambiswas.com/MosamBiswasResume-BW.pdf): black and white, one page - [Résumé PDF, colored](https://www.mosambiswas.com/MosamBiswasResume.pdf): the same content in the site's design, two pages - [Certificates](https://www.mosambiswas.com/certificates.html): eleven certifications and recognitions, from an IEEE conference to Samsung's AI curriculum ## Experience - [Software Developer Intern, Sedna Technologies](https://www.mosambiswas.com/resume.html#experience): Jan / Apr 2026, Mumbai. Shipped work across 4 client projects and built a client-facing dashboard solo (Angular, Flask, MySQL), later submitted as the B.Tech dissertation ## Work - [AMINAL](https://github.com/BiswasMosam/Aminal-Public): local-first voice assistant for Windows. Speech, LLM intent, vision and text-to-speech fully on-device; 29K lines of Python, 125 commands, 849 tests. Python, Whisper, Ollama. 2026. The code is private; the public repo is the write-up - [Locals](https://www.mosambiswas.com/Locals/): a town of AI users that test web and Android apps the way real people do, and say where they got stuck. 114 locals from 9 job families, built from published research and marked draft until people in those jobs confirm them; human limits are enforced in code, not the prompt, and everything runs on a local model. Python, Ollama, LLM agents, Playwright. 2026. Live - [Fill.ai](https://www.mosambiswas.com/Fill.ai/): AI form-filling Chrome extension. Every answer must cite a profile field or it is thrown out; a 121-check end-to-end suite runs in Chrome on every push, and releases go out through CI/CD. JS, Chrome MV3, LLMs. 2026. Source: https://github.com/BiswasMosam/Fill.ai - [MindFirst](https://github.com/BiswasMosam/Mind-First): mental health screening. A GAD-7 style pre-screen run through a Random Forest decides who takes one of 7 condition tests, then hands back self-help and a doctor booking. Next.js, Flask, Prisma. 2025 - [PixelForge](https://github.com/BiswasMosam/PixelForge): leukemia cell classification on C-NMC at 0.89 ROC-AUC, 84% accuracy and 95% recall; a custom spectral-norm GAN augments scarce data. PyTorch. 2025 - [PixelShift](https://www.mosambiswas.com/PixelShift/): in-browser batch image converter, nothing is uploaded. Vanilla JS, Canvas. 2025. Live - [Rail Mitra](https://github.com/BiswasMosam/RailMind): railway platform and track optimizer; constraint programming hands back conflict-free schedules. Python, OR-Tools. 2025 - [IntelliStatement](https://github.com/BiswasMosam/LLM-bank-statement-analyzer): LLM bank-statement analyzer, extraction from PDFs and images into structured data. FastAPI, Streamlit. 2025 - [Committee Connect](https://github.com/BiswasMosam/CommitteesManagement): Android app for college committees, a four-person RAIT mini project synced over Firebase. Android, Java. 2024 - [Sim Sim](https://github.com/BiswasMosam/SimSim-Virtual-Assistant): voice-driven AI desktop assistant with 23 commands, from alarms and news to WhatsApp and games. Python, PySide6. 2024 - [Flex Card](https://github.com/BiswasMosam/Flex-Card): Windows app that scans your hardware and exports a share-ready PNG spec card. Python, PyQt5. 2024 - [To-Do](https://www.mosambiswas.com/To-Do/): full-stack task boards with sign-in, Node backend to front end. Node.js, JS. 2024. Live - [Pops](https://github.com/BiswasMosam/Pops) and [Order Control](https://github.com/BiswasMosam/Order-Control): two offline Android apps for real small businesses, a fabrication firm's order tracker and a food stall order board, on relational SQLite. Flutter, Dart. 2026 ## Proof - [Fill.ai tests](https://github.com/BiswasMosam/Fill.ai/actions/workflows/tests.yml): GitHub Actions runs 44 unit tests and the 121-check end-to-end suite, which loads the real extension into Chrome, on every push. Releases are built by [a second workflow](https://github.com/BiswasMosam/Fill.ai/actions/workflows/publish.yml) - [Locals tests](https://github.com/BiswasMosam/Locals/actions/workflows/tests.yml): 133 tests on every push, with Fill.ai built beside it so its task cards are checked against the real extension - [Dacx tests](https://github.com/BiswasMosam/Dacx/actions/workflows/tests.yml): 60 tests and static analysis on every push, including the connection layer against a real local WebSocket server - [Peghunt tests](https://github.com/BiswasMosam/Peghunt/actions/workflows/tests.yml): 44 tests and static analysis on every push - [Order Control tests](https://github.com/BiswasMosam/Order-Control/actions/workflows/tests.yml): 13 tests and static analysis on every push - [Shellfolio tests](https://github.com/BiswasMosam/Shellfolio/actions/workflows/tests.yml): go vet, the tests and the Linux build on every push. The server itself is live: `ssh mosambiswas.com` - [PixelForge results](https://github.com/BiswasMosam/PixelForge#-results-stage-1): the ROC curve (AUC = 0.89) and the confusion matrix behind the numbers above are in the repository as PDFs, next to the notebook - Live apps anyone can open: [Locals' window on real kept days](https://www.mosambiswas.com/Locals/), [PixelShift](https://www.mosambiswas.com/PixelShift/), [To-Do](https://www.mosambiswas.com/To-Do/), [Fill.ai's latest release](https://github.com/BiswasMosam/Fill.ai/releases/latest) - The IEEE paper below is on IEEE Xplore under its DOI ## Research - [Comparative Analysis of Psychometric Data for Assessing Mental Health Testing Needs Using Machine Learning Models](https://doi.ieeecomputersociety.org/10.1109/ICAIIHI67124.2025.11403838): co-author, IEEE Xplore, Proc. ICAIIHI 2025. Five models benchmarked on 28.5K PHQ-9 / GAD-7 records; Logistic Regression led at 99% accuracy with SHAP explanations - [The paper as a PDF](https://www.mosambiswas.com/ResearchPaper.pdf) ## Elsewhere - [GitHub](https://github.com/BiswasMosam): every public repository - [LinkedIn](https://linkedin.com/in/mosambiswas) - [Google Developer profile](https://g.dev/MosamBiswas): 25+ codelabs, Google Cloud Innovator ## Optional - [SheiChobi](https://www.mosambiswas.com/sheichobi/sheichobi.html): the photography portfolio. Bengali for "that picture"; street, portraits, monochrome, concerts, wildlife and traditions, 84 frames since 2022 - [Site source](https://github.com/BiswasMosam/BiswasMosam.github.io): how this site is built. Hand-written HTML, CSS and JS with no framework or build step, a WebGL ember shader, and an offline service worker - [Shellfolio](https://github.com/BiswasMosam/Shellfolio): the Go server on the bare domain. In a terminal, `curl mosambiswas.com` prints the résumé and `ssh mosambiswas.com` opens the portfolio as an app - [humans.txt](https://www.mosambiswas.com/humans.txt): credits for the site, its typefaces and its tools