Website:
peoplecap.in
Job details:
Our client, a financial services group spanning lending and wealth/asset management, is hiring the single accountable leader for its AI Lab, research, platform, delivery, and governance under one owner, reporting to the CTO.
This is a player-coach mandate. You will set multi-year AI direction and represent AI to the board and regulators and you will also write architecture documents, review critical-path code, and debug production incidents. If your last three years were slide decks and steering committees, this is not your role.
What you will own:
- End-to-end AI delivery across lending and wealth business lines: intake, prioritization, production rollout, adoption, and measured business impact — every delivery with a named accountable engineer, a written success metric, a date, and a kill criterion
- The AI platform: model serving, feature stores, observability, CI/CD for ML, and an internal AI marketplace
- The technical bar: model selection, evaluation methodology, and the internal standard for production-ready AI — including what gets rejected
- An operating model that keeps business-line AI teams autonomous on outcomes but consistent on platform, evaluation standards, security, and risk posture
- Velocity and quality metrics, published and defended: spec-to-production cycle time, eval pass rates, incident rates, drift SLAs, cost per inference, KPI lift per shipped capability
- Risk and responsible AI as delivery gates you own — model risk, data exfiltration, prompt injection, third-party model risk — in partnership with the CISO, without outsourcing accountability
- Build/buy/partner decisions justified by delivery throughput, unit economics, or risk posture
What you need (hands-on, not supervisory; you should be able to whiteboard the system, write the reference implementation, and give substantive PR feedback in each):
- LLM and GenAI systems at production scale: RAG, agentic workflows, fine-tuning (SFT, DPO, LoRA), evaluation harnesses, cost and latency optimization
- Classical ML and quantitative methods: supervised/unsupervised learning, time-series and state-space models, portfolio optimization, factor models
- MLOps and platform engineering: online and batch serving, feature stores, versioning, drift detection, GPU capacity planning, on-prem and hybrid deployment
- Data architecture: lakehouse patterns, contracts and lineage, governance for regulated data
Profile:
- Engineer first: CS, EE, or equivalent quantitative background; a CV that reads as an engineering progression that earned its way into leadership
- 15+ years total, with 7+ in applied ML/AI; last 5+ leading AI/ML engineering teams shipping models into production in regulated or high-stakes environments (financial services, healthcare, telco, large-scale consumer platforms)
- Has survived a production AI failure and can articulate what changed in engineering practice afterwards
- Fluent in trade-offs — cost vs latency, build vs buy, foundation model vs fine-tune vs classical — without retreating into "it depends"
- Writes well: architecture documents, board memos, post-mortems
Click on Apply to know more.