— hey, I'm Rushikesh.
AISoftwareEngineershipping0→1,end-to-end.
- 4+Years Shipping
- 50→1 Products Owned
- 2.2k+Extension Downloads
- Cal.comOpen-Source
AI, frontend, backend — shipped by one person,not a committee.
0→1, end-to-end, mostly caffeinated.
I build AI products the way a startup needs them built — owned end-to-end, shipped without a committee, and treated as a craft rather than a checklist. Over the last four years I've co-founded two SaaS products, embedded an MCP server into VS Code, shipped RAG into real products, and rebuilt legacy codebases into clean, typed, modular systems. Most of my career has been at small teams where one engineer ships an entire feature — frontend, backend, AI, infra. That's the work I love. I have fun with the complex stuff most people avoid.
Today I spend most of my time on AI systems — agent runtimes, Model Context Protocol (MCP) servers, RAG pipelines, LangChain / LangGraph orchestration, and the streaming UIs that turn LLMs from impressive demos into tools people actually use. The frontend and backend stay sharp: modern React, Next.js, TypeScript, Node, Postgres, the boring fundamentals done well.
I've co-founded two SaaS products — Rivaly (AI competitive intelligence) and Onramp (OSS issue discovery in under ten seconds, with a talent surface for recruiters). Both are in active development. Earlier I shipped a VS Code extension with an embedded MCP server that crossed 2,200+ downloads, contributed fixes to Cal.com, and led migrations, billing systems and auth flows as a solo engineer at multiple early-stage startups.
— Rushikesh
AI Software Engineer · Pune, Maharashtra
- 01
Own the feature, end to end
Most teams I join are small. I take a problem from product brief to production — AI, frontend, backend, infra, the boring auth flow, the messy webhook. I prefer fewer hand-offs and more shipping.
- 02
Zero to one, on purpose
0→1 is what I love. Blank page, real users, real constraints, real deadlines. I've co-founded two products and built five from a Figma file to a deploy URL.
- 03
Have fun with the hard part
Streaming, agents, race conditions, weird Stripe edge cases, that one webhook that drops on Tuesdays. The complex stuff is the fun stuff. I look for it on purpose.
- 04
Leave the codebase warmer
Types, contracts, honest comments, small refactors as you pass through. Tools you'd hand to a friend on day one. The next engineer matters as much as today's PR.
A timeline of products I've shipped.
four companies, one consistent thread.
Forge Code
AI Software Engineer · Full Stack
Owned a VS Code extension and embedded MCP server end-to-end — real-time AI agents inside the IDE. Crossed 2,200+ downloads.
- Architected a Model Context Protocol (MCP) server inside a VS Code extension using JSON-RPC 2.0 for real-time AI ↔ IDE communication.
- Engineered an ACP-compliant multi-agent chat interface with low-latency streaming UI and asynchronous, cancellable tool execution.
- Designed REST APIs in Node.js + Express with Redis caching; analytics on PostHog + PostgreSQL aggregations.
- Shipped a scalable frontend with React, Next.js and Zustand — Lighthouse scores 90+.
- Drove the extension past 2,200+ marketplace downloads with zero formal marketing.
Things I've helped put into the world.
a small list, deliberately.
Rivaly
In developmentin progressAn AI competitive-intelligence platform for operators and founders — real-time business insights, alerts and ‘what your rivals shipped this week’ digests powered by LLMs, RAG and scheduled agent workflows.
Onramp
In developmentin progressA two-sided platform built for the open-source generation: developers find skill-matched OSS issues in under 10 seconds via semantic search, and their contributions quietly surface their profile to recruiters hunting for proven, talented engineers.
ForgeCode
ShippedfeaturedA VS Code extension that brings a CLI-based AI agent inside the editor over an MCP server with JSON-RPC streaming. 2,200+ marketplace downloads.
PrepWise AI Interview
Side projectAn AI-powered voice interview simulator using Firebase, Vapi AI and Google Gemini for dynamic question generation.
Cal.com
OSSResolved internationalization bugs and corrected i18next implementation issues across UI components.
PRs #15657 · #15463 · #15626 · #15631
Issue Tracker
Side projectA Jira-like issue tracking system built on Next.js, Prisma, SQL, NextAuth, React Query and Radix UI.
psst — more lives quietly on GitHub.
Browse GitHubThe tools I reach for on Monday morning.
a toolkit, not a religion.
- OpenAI (GPT-4o, o1)
- Anthropic Claude
- Gemini
- LangChain & LangGraph
- LlamaIndex
- Vercel AI SDK
- Model Context Protocol (MCP)
- RAG Pipelines
- Vector DBs · pgvector · Pinecone
- Function Calling & Tool Use
- Structured Outputs
- Embeddings & Reranking
- Evals · LangSmith · Promptfoo
- Agent Workflows
- Streaming UIs
- Prompt Engineering
- TypeScript
- JavaScript (ES6+)
- Python
- Java
- Next.js
- React
- Redux Toolkit
- Zustand
- TanStack Query
- Tailwind CSS
- shadcn/ui
- Zod
- Web Accessibility
- Node.js
- Express
- REST APIs
- Clerk Auth
- RBAC
- Webhooks
- Cron
- Stripe
- Prisma
- Redis
- PostgreSQL
- MySQL
- MongoDB
- Supabase
- AWS
- Vercel
- Docker
- CI/CD
- GitHub Actions
- Jest
- React Testing Library
- Cypress
- API Testing
— the engineering journal
Things I've learned the hard way.
essays, notes, post-mortems.
- 01
From Chains to Graphs: Why LangGraph Is How I Build Agents Now
LatestLinear chains break the moment your agent needs to retry, branch, or wait for a human. Why I migrated from LangChain chains to LangGraph for every multi-step AI workflow — and what production taught me on the way.
AI / GenAILangGraphAgents11 min read - 02
Building an MCP Server Inside VS Code: JSON-RPC at the IDE Layer
A deep dive into bridging AI agents with the editor using the Model Context Protocol, JSON-RPC 2.0 streaming, and a chat surface that finally feels native to the IDE.
AI SystemsVS CodeMCP12 min read - 03
RAG in Production: What the Tutorials Don't Tell You
Chunking strategies, retrieval quality metrics, eval loops, and the unglamorous infra work that decides whether your RAG actually ships — beyond the demo.
AI / LLMRAGInfra14 min read
written between deploys.
Open the journalHave an AI problem worth
solving together?
I'm exploring mid–senior AI / Full-Stack engineering roles and high-leverage contract work — especially anything at the intersection of agents, RAG, real-time interfaces and small, motivated teams shipping fast. Any timezone. Any location.
Reply within a day.
Roles, intros, references.
Open source, daily work.
Notes on AI and craft.
P.S. I read every message.