My magic is never giving up
Lalit Negi

Lalit Negi

Full Stack & AI Engineer

Open to opportunities

GSoC '25 contributor at OWASP.

I build real-time systems, LLM agent pipelines, scalable backends

$20k+ earned in open-source bounties.

Contest rating

1,815

Top 7.77%

10 contestsPeak 1,815Global #67,588

686solved
Easy212 / 974
Medium413 / 2,136
Hard61 / 986
11-day streak148 active days๐Ÿ… 500 Days BadgeMostly Python

GitHub

Contributions and activity

Lalit Negi

GSoC @OWASP

113 repos101 stars38 followers

Open-source bounties

$20k+earned

Paid for merged work on Gumroad & Antiwork repos

0+

Merged contributions

0

Organizations

0

Repositories

GSoC '25

Summer of Code ยท OWASP

GitHub contributions for igennova

Organizations I've shipped to

Tap a logo to filter

Experience

Where I've worked and what I shipped

  1. Oct 2025โ€” Feb 20265 mos

    Oct 2025 โ€” Feb 2026 ยท 5 mos

    Open Source Contributor

    Gumroad / Antiwork โ†—
    5m โ†’ 4.7s135M-row CSV export
    • Optimized a 135M-row CSV export from 5+ minutes down to 4.7s.
    • Migrated legacy Rails pages to Inertia.js with merged PRs to production.
    • Refactored legacy SCSS to Tailwind CSS across the app.
    RailsInertia.jsPostgreSQLTailwind
  2. Jul 2025โ€” Aug 20252 mos

    Jul 2025 โ€” Aug 2025 ยท 2 mos

    AI Engineer

    PUCH AI โ†—
    10k+users rate-limited
    • Built a multi-tool LLM agent pipeline with structured tool calling.
    • Added Redis + Lua rate limiting serving 10k+ users.
    • Unified image and video generation infrastructure.
    LLM AgentsRedisPython
  3. May 2025โ€” Aug 20254 mos

    May 2025 โ€” Aug 2025 ยท 4 mos

    GSoC 2025 Contributor

    OWASP Foundation โ†—
    <150msWebRTC call latency
    • Built real-time video calling with WebRTC at sub-150ms latency.
    • Shipped a WebSocket leaderboard for 100+ concurrent users.
    • Deployed 5+ security simulation labs.
    WebRTCWebSocketsSecurityDjango
  4. Apr 2025โ€” Jun 20253 mos

    Apr 2025 โ€” Jun 2025 ยท 3 mos

    Backend Developer

    Persist Venture โ†—
    Read replicasPostgreSQL scaling
    • Engineered FastAPI backend services and an AI media generation pipeline.
    • Optimized PostgreSQL with indexing and primary-replica read scaling.
    FastAPIPostgreSQLAI Pipeline

Projects

Things I've built and shipped

github.com/igennova/Distributed-Vector-Search-Engine
shard 0HNSWshard 1HNSWshard 2HNSWshard 3HNSWquery ยท top k=10coordinator ยท merge top-K4 shards ยท gRPC ยท parallel fan-out ยท p50 1.6 ms
01ANN search engine, built from scratchIn progress

Distributed Vector Search

A distributed approximate-nearest-neighbor engine exploring the internals of systems like Qdrant and Milvus โ€” HNSW indexing, sharding across gRPC services, and a scatter-gather coordinator, all benchmarked for recall vs latency.

  • HNSW from scratch: 0.93 recall@10, 2.5ร— faster than exact
  • gRPC shards + scatter-gather top-K merge
  • 3.5ร— lower fan-out latency (5.6 โ†’ 1.6 ms) after diagnosing GIL contention
PythongRPCProtocol BuffersNumPypytest
cp-nextjs-iota.vercel.app
LeetFight screenshot
02Competitive coding, as a sport

LeetFight

A gaming-style coding arena where developers face off in real-time DSA battles. Built with Suryansh Singh to make coding social, competitive, and fun.

  • Real-time 1v1 battles
  • ELO rating
  • Leaderboards
  • Private friend rooms
Next.jsTypeScriptWebSocketsReal-time
quotesnap-vfaw.vercel.app
WeMakeVideos screenshot
03Prompt in, finished video out

WeMakeVideos

AI video generation SaaS โ€” turns a single idea into script, voiceover, scene composition, and a rendered edit.

  • Script generation
  • AI voiceover
  • Remotion rendering
Next.jsRemotionTypeScript