AI Engineer

Role Details • Full-Time • Office Location - New York City (hybrid) Why Translucent Healthcare providers drive $2.5 trillion in medical expenditures annually — and operate on razor-thin 2–5% margins. Despite these stakes, the finance teams behind these organizations are buried in spreadsheets, manual data pulls, and disconnected systems, spending more time finding and cleaning data than actually using it to make decisions. Translucent is changing that. We're building the agentic AI platform designed exclusively for healthcare finance— giving every finance team, department, service line their own arsenal of AI Agents that run 24/7, understand their specific data, business logic, and workflows. Founded in 2024 and backed by GV, NEA, FPV, and Virtue, we've already been deployed by healthcare organizations managing over $5 billion in combined revenue. The product-market fit is real, the problem is massive, and we're just getting started. If you want to work at the intersection of AI and one of the most complex, consequential industries in the world, this is the place. Role Overview We're looking for an AI Engineer to help build the agentic platform at the core of Translucent: the AI financial platform transforming how healthcare organizations make business decisions. You'll work hands-on across R&D and platform engineering — improving the underlying agent capabilities every team builds on, rather than configuring agents for a single customer. You'll standardize the tool surfaces new capabilities plug into, make the agent platform ready to integrate across products, and own the evals, context, and harness engineering that make it reliable enough for healthcare finance. You'll match the right agentic architecture to each use case, turn core capabilities into an ecosystem others build on, and own quality end-to-end. If you are energized by fast-paced environments, love sweating the details of production AI systems, partnering closely with product and design, and seeing your work in customers' hands quickly, you'll feel at home here. What You’ll Do • Standardize tool and skill surfaces — define the contracts (MCP-style tool, connector, and skill interfaces) that let us add new capabilities continuously without destabilizing the platform as it grows • Make the agent platform integration-ready — build the connective tissue so agents, tools, context, and data compose cleanly across products, and match the right agentic architecture to each use case • Own evals and benchmarks — build production-grade eval harnesses, replay, and benchmarks for agentic AI, and gate releases on them; in healthcare finance, accuracy is non-negotiable • Lead context and harness engineering — establish best practices for grounding agents in each customer's business rules and data, and for the control loops that keep outputs reliable • Fine-tune and evaluate models — fine-tune in-house and open-source models, benchmark them against frontier baselines, and make the build-vs-buy calls on model, framework, and infra • Turn capabilities into an ecosystem — build reusable core capabilities that translate across features and products, and ship them end-to-end with product and design What You Have • 3+ years of software engineering experience, with meaningful production work in Python • Direct experience shipping LLM-powered or agentic systems in production — not just prototypes. You know where models fail and how to make them reliable • Hands-on experience with at least one modern agent framework (LangGraph, Google ADK, LlamaIndex, Claude Agent SDK, or equivalent), and a clear point of view on what to use vs. build • Production evals and benchmarking — you've built eval harnesses and benchmarks that gate real releases, not one-off spot checks • Context engineering — grounding agents through retrieval, context layers, and knowledge transfer so outputs match a specific org or user • Comfort with the data and retrieval layer — SQL, BigQuery or a comparable warehouse, embeddings, and vector search • A bias toward shipping — you take an ambiguous problem to a working V1 fast, then iterate Plus deep, demonstrable expertise in at least one (ideally two) of the following: • Tool-surface and connector platform engineering — standardized tool/skill contracts (e.g., MCP) and connector ecosystems that scale without breaking • Harness and loop engineering — agent control loops, orchestration, and inference harnesses; right-sizing agent architecture per use case and keeping it reliable at scale • Open-source and in-house model fine-tuning — fine-tuning open models and benchmarking them against frontier baselines • Production context-layer engineering — systems that capture business rules and preferences and feed them to agents reliably Nice to Have • Experience with healthcare finance, accounting, or other structured financial data — or genuine curiosity about it • Applied ML or model-evaluation research background • Open-source contributions to agent or LLM tooling • Experience designing tool-use APIs or developer-facing AI products • Familiarity with our stack: Vertex AI (Gemini, Claude on Vertex), GenKit, MCP, BigQuery Experience Level We're looking for engineers who own meaningful surface area and drive work forward independently. You'll bring sharp judgment on architecture and build-vs-buy trade-offs under ambiguity, a bias toward shipping a working V1 fast, and the craft to make agentic systems reliable enough for healthcare finance — not just demos. https://www.translucent.co/ Base Salary: $175,000-$275,000 USD + equity