What the work involves
This is a builder's seat, not a research post. Day to day you are designing retrieval pipelines — chunking strategies, embedding choices, vector store selection and index tuning — and wiring them into LangChain or equivalent orchestration. Agent and tool-calling workflows sit alongside that: deciding when an agent loop is justified versus a deterministic chain, bounding retries and cost, and handling failure modes when a tool returns garbage. You will also write SQL against the data team's tables, sit in architecture and code reviews, and take business requirements from SMEs and turn them into something that survives production traffic.
Because the role is explicitly tied to roadmap input, expect a meaningful share of time in conversations rather than an editor: scoping what is feasible with current models, pushing back on use cases that retrieval cannot fix, and explaining evaluation results to non-technical stakeholders.
What the platform screens for
- Depth under follow-up. Turing's AI-led screen asks a general question, then drills. Claiming "built a RAG system" invites three questions about your retrieval evaluation and why you chose that chunk size.
- Verifiable specifics. Model names, vector databases, cloud services, latency and cost numbers, team size, what broke.
- Evaluation judgment. How you know a pipeline got better — offline eval sets, human review, regression tests on prompts — not just "it felt more accurate."
- System-level thinking. Deployment, observability, token cost, versioning of prompts and indexes.
Logistics
Full time, based in India, with the 7–12 year experience band applied to your overall engineering career and 1+ years to LLM-specific work. The posting describes cross-functional and possibly client-facing collaboration, so overlapping hours with business stakeholders should be assumed. Compensation is undisclosed in the listing; rates observed on Turing for senior GenAI engineering in India vary widely by client, and nothing here is a guarantee.