Building intelligent systems and scalable digital experiences — from a live pet-care platform with an AI symptom checker, to a flagship realtime messenger, to automation that saves businesses hours every week.
I'm a Computer Science & Engineering student at Northern University Bangladesh who learns by shipping real products. My biggest build, PawMatch, is a live pet-care & adoption platform — realtime messaging and calls, shelter workflows, and an AI symptom checker that helps owners decide how urgently a sick pet needs a vet.
I care about software that is safe by design: the symptom checker's AI can only ever raise the urgency of a case, never lower it — a principle that became the subject of my first research paper, now in preparation for peer-reviewed conference submission.
Alongside my studies I build AI automations with n8n + OpenAI and take freelance work — chatbots, workflow integrations and web apps for small businesses.
The stack I actually build with — from C++ desktop apps to serverless AI products.
Real products, live on the internet — not tutorial clones.
A full pet-care platform: adoption workflows, shelter dashboards, realtime chat & WebRTC calls, and live data sync across devices. Built as a five-person university team project — I owned the healthcare module, messaging, admin panel and authentication.
A species-aware triage engine (29 symptoms, 4 urgency bands) with Gemini photo & video analysis. The AI may escalate urgency but never lower it — a safety-first design I evaluated on an 80-scenario benchmark and am publishing as research.
A flagship-grade realtime messenger rebuilt from research on what users actually want: reactions, pins, polls, stories, scheduled & disappearing messages, and real WebRTC voice/video calls — with a hardened security model.
How do you let a vision model help with a safety-critical decision without ever letting it talk an owner out of an emergency? My paper evaluates a hybrid design — deterministic rules own the baseline, AI observations can only escalate — on a benchmark of 80 guidance-labelled scenarios with a strict development / held-out protocol.
Turning the PawMatch symptom checker into a peer-reviewed conference paper — honest evaluation, held-out validation, and a safety pattern other builders can reuse.
Shipped a five-person team project as a real product on my own domain — realtime platform, AI features, and a growing set of real pets, shelters and adoption applications.
Rebuilt a chat app to flagship standard — full feature set, real WebRTC calling, and a security pass that fixed every finding I could throw at it.
Started offering n8n + OpenAI automation — lead-capture chatbots and workflow integrations for small businesses.
Computer Science & Engineering, Northern University Bangladesh — currently in my second year, Section 5A.
Have a project, a freelance automation job, or a research idea? My inbox is open — I usually reply within a day.