Forward-deployed AI engineer. Shipping software since 2018.
I teach developers and engineering teams to build production AI systems.
Harnesses, agents, evals and guardrails, shipped properly. And the AI-native way of working that lets a small team ship like a big one.
The gap
Anyone can call an LLM API. Shipping AI that works on Monday morning is a different job.
01
It works on the demo. It breaks on real users.
Without evals you are guessing. I teach how to measure agent quality, catch regressions and ship changes with confidence.
02
The bill and the latency arrive later.
Token costs, caching, model routing and timeouts. The engineering that decides whether an AI feature is affordable at scale.
03
Agents with tools need guardrails.
Permissions, prompt-injection defences, human-in-the-loop and audit trails, so an agent can act without becoming a liability.
Work with me
Two ways to learn it properly
For developers
Production AI Engineering
Go from "I've built a chatbot" to designing, evaluating and deploying agentic systems that hold up in production. For working developers with 2–8 years of experience.
Founding batch: waitlist open
For engineering teams
AI-native engineering workshops
Your team already has AI coding tools. Most teams use a fraction of what they can do. Hands-on workshops on agentic workflows, guardrails and measuring the gains.
See the workshop
What I write and teach about
- Harness engineering
- Agents & tool use
- Evals
- RAG that holds up
- Guardrails & security
- Cost & latency
- Deploying on AWS
- AI-native development
Latest writing
All posts →The model is not the product. The harness is.
Why most of the real engineering in an AI system lives in the code around the model, and what that code has to do.
Write the evals before you write the prompt
The single habit that separates AI features that improve over time from ones that silently get worse.
Hi, I'm Shobhan.
I've been shipping software since 2018: full-stack apps, serverless systems on AWS, identity and security platforms. Today I work as a forward-deployed AI engineer, building agentic systems that have to survive real users. I also teach: courses, a weekly newsletter, and short lessons for thousands of developers on Instagram.
More about me →Start with the free email course
Seven short lessons on what separates a production AI system from a demo. Confirm your email and read them right away.