Vraj Shah

Product Engineer

Bengaluru, IN

cheesetosht@gmail.com vrajshah.in github.com/cheesetoshtlinkedin.com/in/vraj-sh

Product engineer specialising in frontend with 6 years at high-growth startups. Built component libraries, server-driven UI systems, and real-time interfaces from scratch. I care deeply about performance, visual craft, and code that scales.

Experience

Product Engineer (Contract)

Wavelength

–

  • Designed and developed a user modeling agent that builds vector images of users from scratch from both product & tech aspects
  • Architected a rudimentary custom memory system for the agents

Product Engineer

Induced AI (Backed by Sam Altman, shut down Dec 2025)

–

  • Worked directly with founder & COO across backend, DevOps, applied AI, and design engineering on multiple client products
  • Oximy (enterprise AI observability): Architected LLM API proxy processing production traffic with <50ms overhead; engineered Redis-based distributed policy engine for sub-millisecond rule evaluation; built real-time analytics dashboard for metrics and logs
  • Built AI platform reducing analyst research from weeks to minutes — structured company profiling, RAG-based agents, long-running workflow harness with step-level traceability across Frontend, Backend, and Worker layers, persistent agent state in MongoDB, and migrated to S3 for horizontal scaling
  • DNTEL: Built dynamic form rendering framework and document processing pipeline
  • Induced Stack: Authored internal reusable package system covering email, auth, and notifications

Software Engineer

Loop Health (YC W20)

–

  • Built fullstack claims support tool saving ~9,000 agent hours monthly reducing context switching dependency on external tools
  • Designed a claim status normalisation layer integrated with 10+ TPAs for accurate claim status tracking, reducing support tickets by 50%
  • Implemented feature flagging and A/B testing infrastructure using Flagsmith, enabling controlled rollouts, experimentation and data-driven product decisions across teams
  • Owned features across 8 different projects along with initiatives in product excellence & developer experience

Founding Engineer

Bluelearn (Shut down Apr 2024)

–

  • First engineering hire, built & led a 5-person team. Established tech stack, coding standards & devops processes from scratch
  • Created custom Tailwind-based headless component library (pre-ShadCN era) used across all frontend products enabling consistent UI with minimal duplication
  • Built real-time chat system from scratch using Socket.IO (5,000+ messages/day) and Redis-backed reactive ranking system for hiring platform
  • Engineered internal services replacing 15L+ in annual external tool costs: IP-based user tracking and custom push notification system with granular SQL-powered targeting
  • Architected server-driven UI system for React Native platform; JSON schema from backend controlled dynamic UI rendering enabling rapid product experimentation without app releases

Projects

Hypersift

–

Real-time communication platform for AI-driven recruitment screening: captures and stores audio-video recordings, streams audio to Deepgram for live transcription, and pipes transcripts through an LLM for candidate evaluation. Built as a freelance engagement.

Technologies: WebRTC, WebSockets, Next.js, Groq, Deepgram, OpenAI

Dooreye

–

Privacy-first neighbourhood security app. Backend in Go + Postgres, React Native frontend with Unistyles.

Technologies: Go, PostgreSQL, React Native

Hackathon Check-in Backend

–

Full-stack check-in system for hackathon participant management, built for HackSVIT

Technologies: Go, PostgreSQL, React Native

Skills

Languages
TypeScript, JavaScript, Go, Python
Frontend
React, Next.js, Svelte, Astro, React Native, Tailwind, shadcn, GSAP, CSS / SCSS, WCAG / accessibility, PWA, Media streaming, Service workers
Backend
Node.js, Express, Hono, GraphQL, FastAPI, Flask, PostgreSQL, MongoDB, Redis, Supabase, WebSockets, REST APIs
AI & LLMs
OpenAI, Groq, Deepgram, RAG, LLM proxies, AI agents, Agentic workflows, Agent memory, Multi-model orchestration, Voice AI, LLM guardrails, AI observability, Exa, Prompt engineering, Structured Outputs, Embeddings / Vector Search, RAG Pipelines, Model Context Protocol (MCP), LLM / Computer Vision Training
Cloud & DevOps
systemd, Nginx / Caddy, Bare Metal Provisioning, Google Cloud Platform (GCP), Kubernetes, AWS, Docker, Vercel, Railway, Heroku, Temporal Workflows, GitHub Actions, CI/CD Pipelines, Serverless, Cloudflare
Architecture & Tooling
Monorepos (Turborepo), System Design, API Design, Microservices, CLI Tools, npm Package Publishing, Claude Code, Vim, A/B Testing, Mixpanel, PostHog, Figma

Education

Volunteer

Co-founder

Hack Club SVIT

–

  • Built a 1,000-member community
  • Directed a national level hackathon (HackSVIT)
  • Led branding, tech, and team of 20