purv kabaria
Purv Kabaria, smiling at the camera

Purv Kabaria

Full-Stack & AI/ML Developer

Location
Surat, India
Email
purvkabaria@gmail.com
Studying
CS @ NIT Surat

CS undergrad at NIT Surat, backend by profession and full-stack by stubbornness. Most of my time goes into AI systems: retrieval pipelines, agentic tooling, and the unglamorous plumbing that makes them fast enough to actually use. I self-host nearly everything I build, mostly because breaking things yourself is the quickest way to understand them.

Currently working with TypeScript, Next.js, FastAPI and PostgreSQL.

Résumé

Experience

  • Stealth AI Startup
    Stealth AI Startup

    Software Developer Intern

    May 2026 - Jul 2026

    • Built an LLM-powered document parsing pipeline using NestJS, Inngest, BullMQ, Redis, and background workers, cutting p99 processing latency from 30s to ~12s while scaling from sequential execution to 100 concurrent document parses.
    • Developed an AI-powered assessment engine, designing backend workflows and LLM integrations for automated candidate evaluation.
    • Engineered third-party integrations with Google Forms, Workday, and Ashby using Firecrawl-based scraping and a GraphQL API layer, auto-generating job descriptions and application forms from source postings in under 30 seconds.
    • Worked across backend systems, asynchronous job processing, and AI-powered product features as part of a 10-engineer founding team at an early-stage stealth startup building a next-generation professional networking and hiring platform.
  • Mercurium Analytics
    Mercurium Analytics

    SDE Intern

    Jun 2025 - Aug 2025

    • Resolved Django ORM N+1 query bottlenecks and nested relationship overhead, and engineered an asynchronous, versioned Redis cache-invalidation system to eliminate partial-state updates, improving response time of data-heavy dashboards by 35%.
    • Built a fault-tolerant data extraction architecture (Python/Selenium) pulling regulatory and financial data from 20+ sources, with independent source isolation, configurable DOM selectors, and exponential backoff retry logic.
    • Architected a high-throughput async pipeline using Redis queues to pass low-overhead AWS S3 metadata pointers rather than full payloads, minimizing message serialization overhead across background workers.
    • Stabilized AWS EC2 node performance under sustained scraping load by enforcing strict browser process lifecycles and structured teardown blocks, eliminating memory leaks.
    • Eliminated double-processing vulnerabilities across distributed background workers by designing atomic database state updates to securely claim processing rights.
    • Refactored frontend components and API endpoints, reducing page load time by 40% and resolving 15+ production issues via Agile workflow.

Leadership & Achievements

  • Chief Executive at Cutting Edge Visionaries (Fintech Club, SVNIT)
  • Developer at ACM SVNIT
  • 12x hackathon finalist, 2x winner
  • Winner, Adani Powermind Hackathon
  • Winner, WebOsmotic AI/ML Hackathon

Tech Stack

React
Next.js
TypeScript
JavaScript
HTML
CSS
Tailwind CSS
Bootstrap
Vue.js
Three.js
Vite
Expo

Competitive Programming

Featured Projects

All projects

IndyChat

Agentic civic-assistant chatbot answering public-safety questions with 95% accuracy, sub-350ms responses, and ElevenLabs voice I/O.

Distributed RAG

Microservice RAG platform with async embedding pipelines, vector search, reranking, and multi-provider LLM orchestration.

Get in Touch

Reach out

I'm always up for collaborating on interesting projects or just chatting about tech. My inbox is open, reach out via your favourite platform!

Usually replies within a day

Open to internships & full-time roles

Send a message

Prefer to write? Drop a note here and it lands straight in my inbox.