AI Product Engineer | Applied AI & Backend Engineering

I build AI-backed products where model output is only one layer of the system: the rest is structured workflows, backend APIs, validation, persistence, human control, testing, and deployment.

My current flagship build is Client Intelligence OS — an evidence-grounded Applied AI system focused on longitudinal client intelligence, accountable follow-up, and human-governed trust.

Currently an AI Applications Development Intern at GreyBox Technologies (Remote, Sep–Dec 2026) and a B.Tech student in Electronics & Communication Engineering at Tezpur University, graduating in 2027.

Portfolio · LinkedIn · All Repositories

Flagship engineering project · Active development · not publicly deployed

Full-stack, multi-tenant client-intelligence system for repeated client/coaching conversations. It turns conversations into source-linked findings, risks, human-reviewed follow-up, and longitudinal client signals while keeping model output advisory rather than authoritative.

• FastAPI + PostgreSQL + SQLAlchemy/Alembic backend with a React + TypeScript workspace
• Human-controlled Action Item workflow with assignments, due dates, completion outcomes, queues, and optimistic concurrency
• Longitudinal signal/evidence/revision model with draft, trusted, rejected, and needs_revalidation states plus What Changed? trajectory surfaces
• Evidence provenance, deterministic factual processing, structured semantic inference, and server-side provider validation
• Workspace-scoped authorization, CSRF/BOLA-oriented access patterns, isolated inference admission, and security-focused regression coverage
• Reviewed design specs, implementation plans, and production/security operating notes kept in-repo for inspectable engineering

Design principle: AI proposes; validated application state and explicit human review remain authoritative.

Live prototype

AI-assisted pharmaceutical complaint-intake prototype that turns text or text-based PDFs into a structured review draft while keeping risk guidance advisory and the final record commit under explicit human control.

• FastAPI + Pydantic API contracts
• LangGraph orchestration with Groq structured output
• PostgreSQL + SQLAlchemy + Alembic persistence
• React + Redux Toolkit review workflow
• Deterministic completeness checks and constrained corrections
• Automated backend/frontend/E2E verification

Live App

Live

Turns exact study scope, readiness, deadlines, and daily capacity into a bounded Study Line, tracks execution, and explicitly replans unfinished work.

• React + TypeScript frontend with FastAPI + Pydantic backend
• Structured LLM outputs with local validation
• Bounded model fallback with deterministic recovery
• Tavily-backed resource ranking
• Topic-scope and daily-capacity invariants enforced in code
• Backend tests plus frontend type/build verification

Live App · Demo

Live

Full-stack portfolio platform for project case studies, technical writing, professional enquiries, and private content administration.

• React + TypeScript + TanStack
• Supabase Auth + PostgreSQL
• Row Level Security and admin authorization
• Versioned database migrations
• Private administration Studio
• Production deployment with SEO/social metadata support

Live Platform

Independent product case + live prototype

Explores a Learn → Build → Earn → Return continuation loop intended to give learners a concrete artifact and reason to return the next calendar day.

• Working static prototype
• Deterministic challenge/scoring logic
• Same-artifact next-day continuation concept
• Explicit experiment metrics and guardrails
• Clear separation between simulated prototype behavior and production requirements

Live Prototype

• AI Internship Match Assistant — Flask-based Resume–JD analysis with file ingestion, match analysis, PDF reports, and honest keyword/phrase-ranked RAG-style retrieval. Live
• AI Health Advisory Assistant — AIoT prototype combining deterministic vitals-risk classification with AI explanation and a demonstrated ESP32/sensor → backend → advisory flow. Live Prototype
• LSTM vs GRU Time-Series Forecasting — Reproducible TensorFlow/Keras experiment with modular Python code, tests, CI, and documented results.
• AI Poetry Chatbot — Flask + Groq application for mood-conditioned poetry generation with PDF export. Live
• InAmigos Foundation — NGO Awareness Webpage — Responsive HTML/CSS internship project. Live

Applied AI: LLM APIs · structured outputs · validation · human-in-the-loop workflows · retrieval-augmented systems · LangGraph

Backend & Data: Python · FastAPI · Flask · REST APIs · Pydantic · PostgreSQL · SQLAlchemy · Alembic · Supabase · Row Level Security

Frontend: React · TypeScript · JavaScript · Redux Toolkit · Vite · HTML · CSS

Reliability & Delivery: pytest · frontend testing · Playwright · Git/GitHub · API boundary validation · deterministic fallbacks · Vercel · Render · Cloudflare

AIoT: ESP32 · sensor integration · MCU-to-backend prototyping · Arduino tooling

• AI Applications Development Intern — GreyBox Technologies · Remote · Sep–Dec 2026
• Web Development Intern — InAmigos Foundation · Remote · Jun–Jul 2026
• B.Tech, Electronics & Communication Engineering — Tezpur University · Expected 2027

I am especially interested in engineering roles where Applied AI, backend systems, product thinking, and reliable human-controlled workflows meet.