- Singapore, Singapore +2 more
- Lead
- 7+ years
Own a multi-agent research system as production infrastructure: make it faster, cheaper, measurably better and efficient.
Locations (3)
- Singapore, Singapore
- Hong Kong, Hong Kong, Hong Kong
- United States
Responsibilities
- Keep the LLM layer reliable: latency budgets, concurrency and queueing, declared degradation modes, structured-output validation and provider failover.
- Build evaluation tied to research outcomes: offline sets from real queries with outcome labels, online experiments, regression gates on every change, and model-generated confidence calibrated with quantitative research before it reaches a decision.
- Control cost and routing: per-workload cost attribution, difficulty-based routing between small and frontier models, caching and batching, and the economics of self-hosted serving and GPU capacity.