Self-taught developer from India, now architecting the systems I used to read about. I work end to end — the retrieval layer, the agent, the pipeline behind it, and the dashboard a human supervises it from. Most of what I ship talks to real customers on the first day.
You describe the workflow; I tell you where the automation actually belongs, what it will cost you to run, and what will break first. Recorded, with notes afterwards.
For teams already building with LLMs. I read the repo, poke the retrieval layer, and give you a written list of the decisions that will hurt in six months — ordered by how much.
A standing seat in your engineering decisions: weekly call, async review, and someone accountable for the AI parts of your roadmap without hiring for it yet.
Agents, retrieval and what production actually demands — taught on your codebase and your data, not on a toy example.
A multi-tenant WhatsApp sales-automation platform: an AI sales agent and a full lead pipeline running on each business's own WhatsApp Business number, in five languages, with a per-tenant retrieval layer so it answers only from that business's own content.
See the product →An AI co-pilot that turns customer reviews into actionable product insight for e-commerce sellers. I am leading the MVP: the ingestion, the clustering, and the part that tells a seller what to change rather than what people said.
Forty-five minutes, your actual workflow, and an honest read on whether AI is the answer or whether you have a process problem wearing a technology costume.