2025 – nowSole engineer
Nexus
Mesa's all-in-one school app
Replaced Moodle with one app for LMS, community, AI interview prep and notifications.
Next.js / Express 5 / Kysely / Postgres / GCP
Case study Live
Sheet 01 — Idea01 / 03
I ship AIproducts, fromschema toproduction.
Name
Gaurav Madan
Role
Forward deployed engineer
Based
Bengaluru, India · IST
Status
Taking new projects
Full-stack AI engineer, 2.5+ years. I embed with founders, scope the MVP, then build, ship and run it in production: agents, web, mobile and cloud.
01By the numbers
Numbers I can defend in a call
- N.01to ship Nexus's LMS, as the sole engineer
- 20 days
- N.02daily active users on Nexus
- 600+
- N.03API responses, down from 3–4 s
- 200 ms
- N.04sign-ups on Wave AI, Android + iOS
- 2,000+
02Selected work
Projects I owned end to end
2026Consulting · end to end
Lumen
WhatsApp AI agent for brokers
Real-estate brokers run their whole workflow on WhatsApp, by text or voice.
Node.js / LangGraph / Gemini / BullMQ / Postgres
2026Consulting · end to end
GTM Suite
AI go-to-market platform
Creator discovery and Reddit lead finding for startups, from a one-line brief.
Next.js / Express 5 / Postgres / OpenRouter / Langfuse
Case study Live
2025Founding engineer
Wave AI
AI companion app
An agentic AI best friend, set up from scratch and launched on Android and iOS.
React Native / Expo / LangChain / Firebase / Mixpanel
2025Lead engineer
Founder's Compass
AI startup mentor
Takes a founder from a raw idea to a scored verdict, then plans the first sprint.
FastAPI / Gemini / Qdrant / MongoDB / Next.js
Case study Live
03Also built
Smaller things, still running
- The AI CTOnpx CLI that helps non-technical founders ship with AI coding agents
- MSL Portalfull-stack operations tracker for Mesa Startup Lab
- Finance dashboard + n8nreplaced manual Excel work; weeks to minutes
- Startup buildsed-tech revenue ₹7–8L → ₹45–50L/mo; ad-tech product to a ₹40L pre-seed
- DorWorldNext.js + GSAP site with Juspay payments
- DorTV notificationsWhatsApp + SMS payment and subscription reminders
04Stack by layer
The stack I actually run in production
- L1Agents
- LangChainLangGraphLangSmithTool callingRAGText-to-SQLMCP
- L2Models & vectors
- Gemini (Vertex AI)Azure OpenAIOpenAIQdrantpgvector
- L3Backend & data
- Node.jsExpressFastAPIPostgreSQLKyselyPrismaMongoDBRedis
- L4Frontend & mobile
- ReactNext.jsTanStack QueryTailwindGSAPReact NativeExpo
- L5Cloud
- GCP Cloud RunApp EngineCloud BuildAWS EC2 / S3 / SESFirebaseDocker
05Experience
2.5 years, four companies, one direction
- Founding engineer for incubated startups across ed-tech, ad-tech and women-only ride-hailing.
- Built Nexus, Mesa's all-in-one app, and its AI agent and MCP server.
- Ran GCP and AWS infrastructure and CI/CD for Mesa and its startups.
- Led web development for DorWorld: Next.js, GSAP, Juspay payments.
- Backend APIs for DorTV (25+ OTT apps, 300+ live channels) and its notification engine.
- Next.js e-commerce with Apollo Client, GraphQL Codegen and SSR.
- MyFatoorah payments for Kuwait; Strapi blog.
- React Native app for Kirana stores with CleverTap analytics.
- A chatbot with Kaleyra SMS login and an S3 sales-report tool.
06How I work
Four steps, from idea to a product that runs
- 01
Embed
I sit with the founders and the people who'll use it, until I understand the job.
- 02
Scope
We cut to the smallest product that proves the idea, and agree on what done means.
- 03
Ship
Weekly releases to real users. Typed, tested, observable from the first deploy.
- 04
Run
I stay on after launch: infra, costs, on-call, and the next iteration.
07FAQ
Questions I get before the first call
If yours isn't here, ask the assistant or write to me. I answer within a day.
Both. I'm happiest owning a product end to end; on teams I take the hardest piece, usually the AI or the backend.