About the Role
We’re seeking a Senior Lead Machine Learning Engineer to architect, ship, and scale the next generation of agentic intelligence across Upwork. You will lead end‑to‑end development of AI agents and the platform that powers them—from LLM training and evaluation to runtime orchestration, safety, and developer APIs. This is a hands‑on, high‑impact role at the intersection of applied research and platform engineering, enabling internal teams and external developers to build reliable, safe, and high‑performing agents on Upwork.
Responsibilities
- Build Agentic Intelligence. Design and implement multi‑agent systems (planning, tool‑use, memory, debate/critique, reflection) with robust guardrails and recovery strategies.
- Develop protocol‑aware agents and services that interoperate cleanly with developer tooling (e.g., agent frameworks and protocols such as MCP).
- Own reliability at scale: deterministic execution where needed, idempotency, timeouts/retries, and evaluation‑driven iteration on agent behavior.
- Train, Align, and Evaluate LLMs for Agents. Lead data strategy and curation for agent tasks; drive SFT, DPO, RLHF/RLAIF, and safety tuning tailored to multi‑tool, multi‑step workflows.
- Stand up evaluation harnesses for functional, task, and longitudinal metrics (success rate, time‑to‑completion, hallucination/escape rates, cost/latency).
- Build policy‑driven guardrails; partner with Legal/Security on data governance and privacy.
- Engineer Agentic Platform Backend Infrastructure. Architect low‑latency inference, retrieval, and orchestration services (streaming, event‑driven pipelines; scalable queues; caching; batching) with strong SLOs.
- Ship production‑grade services (APIs/SDKs, auth, rate limiting, observability) that make agent features easy to integrate for internal and external developers.
- Optimize cost/performance via quantization, distillation, model‑routing, and autoscaling; integrate evaluation signals directly into runtime and CI/CD.
- Lead, Partner, and Uplevel the Ecosystem. Provide technical leadership across research, product, and platform teams; mentor senior ICs; influence roadmaps with clear metrics and trade‑offs.
- Publish internal guidance and exemplar implementations; contribute to technical content, samples, and reference architectures for our agent platform.
- Define and track KPIs for data/quality/throughput, and drive continuous improvement using experiment results and production telemetry.
What it takes to catch our eye
- 8–12+ years in applied ML/ML systems with 4+ years building LLM‑powered products; proven delivery of agentic workflows in production.
- Hands‑on mastery of LLM adaptation (prompting, tool/function calling), data curation, and safety/guardrails.
- Strong software fundamentals (distributed systems, transactions, consistency, resiliency) and experience building high‑throughput microservices/APIs/SDKs.
- Fluency with Python; proficiency in one of Go/Java/Javascript a plus. Experience with container orchestration, messaging/streaming, and observability stacks.
- Experience designing eval suites for agents (task/rubric‑based, offline/online) and closing the loop from evals → training → runtime policy.
- Comfort with cost, latency, and reliability trade‑offs; you use metrics to make crisp decisions under ambiguity.
- Familiarity with agent frameworks and protocols (e.g., MCP; API/SDK design for developer productivity).
- Track record of leading cross‑functional initiatives and mentoring senior engineers; excellent written communication and bias for measurable results.