AI Engineer

Why This Role Exists

The core of the product is a pipeline that reads dense technical drawings, often scanned and inconsistent, into structured data that enterprise systems consume. A page can carry hundreds of tags, and the relationships between them matter more than the tags. Our commitment to a customer is an accuracy number, so a quiet regression is a commercial incident. This role owns that pipeline, and the evaluation system that keeps it honest.

What You Will Own

• The extraction pipeline: vision and language model orchestration, tag and symbol recognition, relationship extraction, cross-document reconciliation

• The full AI development lifecycle: data curation, evaluation design, model selection, deployment, monitoring, and the loop from failures back to improvements

• Evaluation as a discipline: hand-verified golden sets, per-class precision and recall, gated in CI

• Model economics: which model, at what effort, at what cost and latency per document

Requirements & Capabilities

• Good with the AI development lifecycle end to end, in production rather than in demos

• 3+ years building LLM or ML systems with real users

• Strong Python: async, typing, profiling, maintainable code

• Retrieval and context engineering in practice, including knowing when retrieval is the wrong tool

• Evaluation discipline (the strongest signal in an application)

• Document, vision, or multimodal experience

• Nice to have: CV fundamentals, agentic systems, graph or hybrid retrieval

This Role Is Not For You If

• Your experience is assembling frontier-model API calls into a wrapper

• You evaluate by forming an impression rather than measuring

• You want a research role (this is production engineering with a research component)

How We Work

• Remote-first with a Gurugram base

• Small surface, deep ownership (systems, not tickets)

• AI-native (judged on what ships and whether it is correct)

• Real users from week one (feedback reaches you directly)