Skip to content

PROFILE.md

Latest commit

 

History

History
70 lines (54 loc) · 7.3 KB

PROFILE.md

File metadata and controls

70 lines (54 loc) · 7.3 KB

PROFILE.md — Asmit Datta

Quick Pitch

Full-stack engineer who builds products end-to-end — from 3D interactive frontends to GPU-accelerated ML backends — with a knack for identifying the real problem before writing a line of code. Graduating May 2026 from Arizona State University (B.S. Computer Science, Business Minor).

Core Domains

  1. Full-Stack Development (React, Node.js, FastAPI, Go) — production apps across web, mobile, and desktop with polyglot backends
  2. AI/ML Engineering (RAG pipelines, LLM integration, NLP, PyTorch) — voice-to-SQL systems, NER pipelines, semantic search with vector DBs
  3. Real-Time Systems (WebSocket, streaming, live tracking) — bidirectional communication, live dashboards, transport tracking
  4. Mobile Development (React Native, iOS, cross-platform) — HealthKit integrations, Expo, Clerk Auth, production App Store workflows
  5. Computer Vision (YOLOv8, pose estimation, GPU optimization) — multi-model parallel inference at 30+ FPS on HD streams
  6. Cloud & DevOps (AWS, Docker, CI/CD) — S3/SQS/DynamoDB, Cloudflare tunnels, Nginx, containerized deployments

Top Skills (for keyword matching)

TypeScript, JavaScript, Python, SQL, Go, Java, C++, Rust, React, React Native, Tailwind CSS, Three.js, Node.js, Express.js, FastAPI, Flask, PostgreSQL, MongoDB, Firebase, AWS (EC2, S3, Lambda, SQS, DynamoDB), Docker, Kubernetes, WebSocket, GraphQL, LLM APIs, RAG, Pinecone, PyTorch, TensorFlow, spaCy, YOLOv8, OpenCV, Whisper, Electron, Mapbox, Google Maps API, OAuth 2.0, JWT, Mixpanel, DataDog, Redis, Socket.io, Material-UI, GSAP, Howler.js

What I'm Looking For

  • Role types: SWE New Grad, Full-Stack Engineer, AI/ML Engineer
  • Company size: Open — startup pace with ownership, or mid-size/big tech with strong engineering culture
  • Industries of interest: AI/ML tooling, developer tools, health tech, fintech, real-time systems
  • Needs OPT/H1B sponsorship (F1 visa)
  • Available: June 2026 after graduation
  • Location: Open to remote, hybrid, or on-site

Unique Angles / Talking Points

End-to-End Product Builder

Every project in this portfolio was taken from idea to deployed product, not just a code exercise. Pash (audio stem separator) required designing the user-facing React app, the FastAPI + Demucs ML backend, AWS infrastructure (S3, SQS, DynamoDB), OAuth authentication, GPU worker orchestration, and deployment automation — all as a single developer. The portfolio site itself (asmit.space) is a custom 3D interactive environment built with Three.js, complete with character movement physics, raycasting collision detection, a desktop-metaphor file browser, and an integrated full-stack audio app — demonstrating that "full-stack" means the entire surface area, not just a CRUD app with a React frontend.

Problem Identifier, Not Just Problem Solver

At My Yoga Teacher, the challenge wasn't "build a voice recording UI" — it was "business users spend hours writing SQL to pull student health reports." The solution was a voice-to-SQL pipeline (Whisper ASR -> rule-based NLU with 9 intent patterns -> dynamic parameterized query builder) that turned a multi-minute manual process into a spoken sentence. At Nexus AI, customer support agents were drowning in repetitive policy questions — the answer was a RAG pipeline with 500+ embedded policy documents, semantic search via Pinecone, and a fine-tuned BERT intent classifier (95% accuracy across 7 categories) that grounded every response in actual company policy. The pattern: understand the human workflow bottleneck first, then engineer the right system to remove it.

Polyglot Systems Thinker

Comfortable designing systems where the frontend is TypeScript/React, the API gateway is Go, the ML service is Python/FastAPI, and the database is PostgreSQL — all communicating cleanly through well-defined interfaces. The MYT NLP project ran exactly this stack in production with sub-second response times. This isn't about collecting languages; it's about choosing the right tool for each layer and making them work together.

Ships Under Constraints

Built SentinalAI for the Voxel51 hackathon — real-time multi-model computer vision (YOLOv8 + pose estimation + facial detection) running in parallel on a single GPU at 30+ FPS, with privacy-first edge computing and zero cloud dependencies. Built Lynti Mobility to solve a real transportation problem in Northeast India with live tracking, route optimization, and real-time WebSocket updates. These aren't toy demos — they solve real problems under real constraints (time, hardware, connectivity).

Personal Touch

  • The portfolio site has a 3D room, and a desktop interface with draggable windows — because engineering should be fun, and how you present work says something about how you think.
  • Active in ASU's AI Society.

What I DON'T Want

  • Roles that are purely maintenance/ops with no greenfield building
  • Environments where engineers don't talk to users or understand the problem space
  • "AI" roles that are just prompt engineering with no systems work
  • Teams that treat shipping as someone else's job

Resume Highlights (auto-extracted)

Experience

  • My Yoga Teacher — AI Intern (May 2025 - Nov 2025): Built end-to-end voice-to-SQL reporting system (React 19, Go, Python FastAPI, PostgreSQL, Whisper). Designed React Native health metrics dashboard integrating HealthKit and ClickHouse. Full microservices architecture with clean separation across a polyglot system.
  • WatchDNA — Full-Stack Developer (2023): Mapbox-powered asset tracking maps, web scraping with Python/Selenium, JWT auth, 30% faster asset location. React + Node.js + MongoDB stack.
  • Nexus AI Solutions — AI Engineer Intern (Feb - Nov 2024): RAG pipeline with Pinecone/GPT and 500+ embedded policy documents. Fine-tuned BERT intent classifier (95% accuracy, 7 categories). Real-time chat with Socket.io. NER pipeline with spaCy 3.x. Production deployment with Docker, Redis, Gunicorn, DataDog APM.
  • Flickmatch — Intern (Jan - Jul 2023): Marketing campaigns boosting engagement 60%, UI/UX improvements driving 35% user retention increase.
  • CovRelief — Volunteer (May - Sep 2020): Built vaccination registration system, managed data analytics for 15+ students, raised $2,400+ for education.

Projects

  • pash — Audio stem isolation (React, FastAPI, Demucs v4, AWS S3/SQS/DynamoDB, CUDA GPU workers, OAuth 2.0, user management with rate limits)
  • SentinalAI — Privacy-focused computer vision system (YOLOv8, PyTorch, FastAPI, WebSocket, PostgreSQL, Docker, 30+ FPS on HD, edge computing, zero cloud dependencies)
  • Lynti Mobility — Real-time transport app for Northeast India (React Native, Expo, TypeScript, Google Maps, WebSocket, Node.js, PostgreSQL, Clerk Auth, Zustand)
  • NeuroPilot — Desktop AI assistant (Electron, Browser APIs, WebSocket, NLP, OAuth 2.0, IPC)
  • JobifyAI — Job application automation (Python, Selenium, GPT-4, threading) — reduced application time by 85%
  • asmit.space — 3D interactive portfolio (Three.js, Express.js, GSAP, Howler.js, Firebase) with custom physics, collision detection, and integrated full-stack apps

Education

  • B.S. Computer Science (Business Minor) — Arizona State University — May 2026

Research

  • Research Assistant, University of Delhi (Jan - May 2024): ML models for heart disease risk prediction, 86.96% accuracy