AI Backend Engineer · MLOps Enthusiast | B.Tech CSE (AI & DS) '29 @ Assam Down Town University, Guwahati
I build and ship AI-powered backend systems — not just notebooks. Currently focused on audio AI, LLM-powered APIs, and closing the gap between "vibe-coded" and production-grade.
🎯 2026 Internship Goal · 📍 Guwahati, Assam · 🌐 Originally from Tripura
VoiceGuard AI — Deepfake Audio Detection API
- Tech: Python, FastAPI, Meta Wav2Vec2-Large-XLSR-53, Hugging Face Spaces
- Built for the HCL GUVI Buildathon. OWASP-hardened FastAPI backend with JWT auth, CORS, and
.envhygiene. ~91% detection accuracy. Currently refactoring the synchronous inference path (run_in_executor) into a modular routers/schemas/services layout.
MediSense AI — AI Symptom Analyzer
- Tech: Python, FastAPI, Google Gemini 2.5 Flash API, Render, Vercel
- Returns structured condition cards with probability scores and urgency-level triage. Original version was AI-assisted end-to-end; currently being rebuilt from scratch under a self-imposed "Teacher Mode" protocol — AI explains, I implement.
AudioGuard Agent — GitAgent Hackathon
- Autonomous audio-processing agent submission, built in the Antigravity IDE.
- GirlScript Summer of Code (GSSoC) 2026 — Selected Contributor, AI Agents Track + Open Source Track (Apr 2026–present)
- Open Source Connect Global (OSCG) 2026 — Contributor; merged PR into the Intervyo repository, certified
- GDG On Campus, ADTU — Member
Languages
AI / ML
Backend & Infra
Currently closing: async FastAPI (ASGI lifecycles), message brokers (Redis/Celery), WebSocket streaming, DB connection pooling, multi-stage Docker builds
- 🏢 Secure an AI Backend / MLOps internship
- 🔨 Rebuild MediSense AI independently — no vibe-coding
- 🌍 Keep shipping PRs through GSSoC and OSCG
- 📜 Complete IBM AI/DS + Azure MLOps certifications
- 🐳 Move VoiceGuard and MediSense from "deployed once" to containerized, CI-driven workflows