Location: On-site, Bengaluru
Type: Full-time
You live in the modern automation stack: n8n (or similar orchestration), LLM APIs and local models (Ollama and the like), scrapers, and third-party tools wired together into systems that do judgment-heavy work at scale. You're not a pure developer - but you write enough code to break past what no-code tools can do, and you understand why a client needs a given workflow, not just how to plumb the nodes.
The non-negotiable: you take a workflow from "works in a demo" to "a client relies on it." That means error handling, versioning, monitoring, and being the person who fixes it when it breaks at the worst possible time.
-
Productisation: turn a workflow that works for one client into something we sell to ten - documented, versioned, reliable, with sane error handling and guardrails.
-
LLM engineering: prompt design, chaining, model selection (when to use a hosted API vs. a local model), managing cost and latency, and handling the fact that LLM outputs are non-deterministic.
-
Monitoring & maintenance: watch your workflows in production, catch failures, debug and fix them yourself.
-
Iteration: optimise for output quality, speed, and cost over time.
-
Experience in building automation workflows - things that actually ran, not demos.
Be ready to walk us through one that broke in production and how you fixed it. -
You've shipped something a client or external user genuinely depended on � not just internal experiments.
-
Hands-on with workflow orchestration (n8n or equivalent) and LLM-in-the-loop systems - prompt chaining, structured outputs, retries, fallbacks.
-
Experience with LLM APIs and local model setups (Ollama or similar), and a real opinion on when to use which.
-
Comfortable writing code and wiring up APIs / third-party tools when the platform alone won't cut it.
-
Enough marketing literacy to judge whether the output is actually good
-
High ownership. You don't hand off a half-working workflow and call it done.
-
Comfort handling non-deterministic LLM behaviour in production (evals, guardrails, output validation).
-
You can explain a technical workflow to a non-technical client without losing them.
A practical build exercise (architect an LLM workflow from a loose brief and show how you'd make it production-ready) plus a conversation on judgment calls - what you do when an LLM workflow produces garbage output mid-client-engagement, or when a workflow fails and a client's pipeline is on the line.
