Ex-operator, current builder. I have shipped AI into production for SaaS, media and e-commerce.
I am Himanshu Taneja, based in London, UK. I spent a decade inside product and engineering teams before going independent in 2024. Everything on this site is work I owned, mostly from the first line of code to the post-launch retro.
Boring systems. Opinionated agents. Ruthless evals.
My bias is toward the smallest system that could possibly work. I'd rather ship a narrow agent with a reversible action pattern than a general-purpose demo. Evals are how I sleep at night — every production system I've shipped has a regression harness that gates releases.
Opinions are cheap; cost, latency and observability are not. I budget every bet in tokens, wall-time and on-call risk before the first PR lands.
- Weekly written updates with diffs, metrics and open risks.
- Small PRs, merged often, behind feature flags.
- Every customer-facing change passes an eval gate.
- No demoware: if it's not on main, it doesn't exist.
- One long async post per week over a meeting, by default.
- Quarterly exec readouts with $ moved, not screenshots.
A decade of shipping — five years of it with AI in production.
- 2026Fractional AI Lead, two Series A SaaS
Owning AI roadmap and hiring. Shipping agentic systems that pay for themselves inside the quarter.
- 2025Principal consultant, RAG + voice
Led production RAG rollouts and Vapi voice-agent platforms across UK/EU SaaS and media clients.
- 2024Founding AI engineer, ops automation
Built an agentic SaaS from zero to $410k ARR in seven months. Shipped billing, auth, evals and observability.
- 2023Head of Engineering, e-commerce
Ran a 14-person team, put the first agents behind the ops desk, and cut ticket volume by two thirds.
- 2021Staff engineer, data platforms
Led a five-year replatforming: warehouses, ML pipelines, and the first internal LLM tooling.