Full-stack engineer who ships production-shaped systems, not tutorial demos. I write the "why," not just the "what" — every project below has an architecture decision behind it that I can defend in detail.
Currently: Frontend engineer at Sparrowhost, working on likelife.ai (AI voice agent SaaS) and simuphish.com (phishing-awareness training platform).
Building in public: SaaS side-projects that solve real infra problems — webhook reliability, RAG search, video pipelines, link management — each one written up with the trade-offs spelled out, not just the feature list.
📫 vedantbhavsar.10@gmail.com · LinkedIn · X/Twitter · vedxnt.com
| Project | What it does | Stack |
|---|---|---|
| HookScope | Multi-tenant webhook observability platform — verifies, archives, forwards, and audits webhooks with replay and live alerts (private) | Fastify, Express, Next.js, BullMQ, Redis, S3, Prisma, Turborepo |
| DocSense | RAG platform — upload PDFs, get semantic search + streamed LLM answers over your own documents | Next.js, Express, BullMQ, pgvector, Drizzle, SSE |
| ClipFlow | SaaS for YouTube creators — automates scheduling, thumbnails, and chapter timestamps from a single upload (private) | Next.js, Express, BullMQ worker, Prisma, S3/R2, YouTube Data API |
| Sniplink | URL shortener with click analytics, geo-tracking, and Stripe billing, built for sub-millisecond redirects | Bun, Express, React 19, Redis, PostgreSQL (Neon), Prisma |
Each repo's README documents the actual architecture — sequence diagrams, state machines, and a "why this and not that" section comparing the alternative I didn't pick. I'd rather you read those than a list of badges.
- Decouple the hot path from the slow path. Sniplink's redirect endpoint never blocks on a DB write; HookScope's webhook intake never waits on auth/CRUD traffic. Same idea, different repo.
- Durability has a cost — name it. I use Redis as a queue when the infra is already there and the failure mode (a flushed cache before ack) is acceptable; I reach for Postgres-backed queues when it isn't.
- State machines over status flags. ClipFlow's video pipeline is
UPLOADED → READY/SCHEDULED → PUBLISHING → PUBLISHED/PUBLISH_FAILED, with every transition's failure path written down before I wrote the happy path. - Auth that survives a real attacker. httpOnly cookies over localStorage, JWT refresh rotation with theft detection, raw-body capture for HMAC verification before the body parser can mangle it.
Core: TypeScript · Next.js · NestJS · Node.js · Express
Data: PostgreSQL · Drizzle · Prisma · pgvector · Redis
Infra: Docker · AWS S3 · BullMQ · Turborepo · GitHub Actions
Frontend: React 19 · Tailwind · shadcn/ui · TanStack Query · Zustand
Tracking crypto markets with technical analysis, deep in body-recomposition training (vegetarian, gym 5–6x/week), and posting build logs on X as these projects ship.
Open to remote full-stack roles at early-stage product startups. Reach out if any of the above is relevant to what you're building.