THE FULL STACK
(new delhi. no vc. no team. just tabs.)
I'm Ayush Thukral — full stack dev from New Delhi. I build AI-powered products, automate things that shouldn't be automated, and occasionally ship at 3am because a deadline is a vibe.
My AI agents Chotu and Motu run a real company over WhatsApp — a $4.5M sales pipeline included. Chotu does the ops (sourcing, quotes, CRM, 50+ requests a day), Motu does the thinking (strategy, priorities, executive judgment). The client calls them their best employees. They cannot unionize.
I freelanced for Finease — built their entire CRM from scratch. For WellnessZ I shipped an AI meal recs engine + WhatsApp chatbot that cut their manual support by 80%.
Also: writing HiveMind, an operating system in Rust, from scratch, built for AI agents. It boots in 16MB and comes with a research paper. Because Node.js wasn't low-level enough for a Tuesday evening. Pray for me.
RECEIPTS:
2024 — Present
CHOTU & MOTU — THE AGENTS THAT REPLACED AN OPS TEAM
Built two AI agents on OpenClaw. Chotu runs operations over WhatsApp — sourcing, quotes, CRM, a $4.5M pipeline, 123 scripts, 65+ workflows. Motu handles strategy, priorities, and executive judgment. Cut manual admin time by 60–80%. Client reports no desire to hire humans again. I feel slightly responsible.
2024
SOURCEWITHAI.COM — RAG IN PROD
Built the entire B2B sourcing platform. RAG pipeline, vector search, Redis caching, Stripe billing, n8n automation. Live in production. Wrote every line from scratch. Not a template.
2023 — 2024
WELLNESSZ — AI MEALS + WHATSAPP CHAOS
AI meal recommendations (35% engagement boost), WhatsApp Cloud API bot (80% query automation), Razorpay integration (22% checkout improvement), 40% API latency reduction. 1 year in production. Numbers don't lie.
2024
#2 — STELLAR WEB3 HACKATHON
AI arbitrage bot on Stellar blockchain. ML price prediction, smart contract execution, real-time DEX price feeds. Won 2nd. Built in 48 hours. I maintain we deserved 1st.
Ongoing (Tuesday evenings)
HIVEMIND — AN OS BUILT FOR AI AGENTS
Writing an operating system from scratch in Rust — no Linux, no stdlib. Each AI agent gets its own hardware-isolated VM booting in 16MB (a Linux guest needs 128–192MB), all coordinating through one shared memory graph. A local LLM is wired in like a GPU. There's a red-team harness and a research paper in progress. Started because I ran out of interesting weekend projects. It escalated.