• Design, build, and harden agentic workflows that plan and take actions reliably — and understand why agents fail (context loss, compounding errors, no feedback signal) and how to structure tasks so they succeed. • Build retrieval (RAG/search) pipelines that fetch the right client data and ground model outputs in it, integrated into core applications rather than demos. • Select the right model for each task — reasoning model vs. fast instruction model — and be able to justify the trade-off in latency, cost, and quality. • Engineer prompts and context structures appropriate to the model class, including knowing when reasoning models need framing rather than step-by-step hand-holding. • Write evaluations for AI features, because with non-deterministic models “it worked once” is not evidence that it works. • Connect AI tools to internal systems and data sources via APIs or MCP to power real client use cases. • Review and validate AI-generated code and automated workflows critically for correctness, security, and safety. • Collaborate with the Builder and the Integration & Data Engineer to deliver complete, working solutions, and document workflows, prompts, and integrations in Notion.

• Build agent harnesses in Python using LangChain and LangGraph, including tool-calling, structured outputs (Pydantic/JSON schema), retries, streaming, and memory • Package agent harnesses for the AgentCore Runtime with appropriate context, tools, skills, and subagents that fit cleanly into production flows and scenarios • Write the tools and skills agents use: API integrations, SQL queries against Snowflake, Snowflake backed knowledge retrieval with clear contracts and Pydantic validation • Build evaluation harnesses (golden datasets, LLM-as-judge, regression suites) using AgentCore Evaluations, and wire them into CI • Implement guardrails around tool execution: auth scoping, input/output validation, PII and prompt-injection protections, and hallucination mitigation • You own what you ship: prototype, deploy through Amazon AgentCore, monitor traces, and fix it when it breaks • Partner with data engineers on Snowflake backed retrieval patterns (Cortex Analyst and Cortex Search Services) • Contribute to refining our internal engineering patterns as the stack evolves

1 in 4 people in the US have a treatable mental health condition, but most providers don't accept insurance, making therapy too expensive for most people. Headway’s mission is to fix this by building a new mental healthcare system everyone can access. We started by solving the biggest barrier to care: insurance. The admin work - credentialing, claims, payment reconciliation - is a nightmare. We've automated that. But we're going further. Over 75,000 providers across all 50 states run their practice on our software, serving over 1 million patients. We are building the best tools for therapists to run their entire practice, reimagining the experience of finding a therapist, and investing in the platform foundations to enable this at scale. We aren't just a billing layer; we are becoming the platform where care actually happens. We're a Series D company with $325M+ in funding (a16z, Accel, Spark Capital, etc.), looking for exceptional people to help us achieve this mission. We want your time here to be the most meaningful experience of your career. Join us, and help change mental healthcare for the better. About This RoleYou will be Headway's first AI Platform product manager. Your mission: build the AI platform that makes every Headwegian an AI-native employee. In Q1 2026, we launched Eddy — an internal AI assistant with direct access to Snowflake, Jira, Slack, GitHub, Google Workspace, and more. Early adoption is strong in R&D, but usage drops off sharply outside of it. Eddy is the first of what we believe will be many: an ecosystem of agents, skills, evals, and data integrations that together become Headway's AI platform layer. You'll own that vision — and the roadmap to get there. This is not a classic internal tools role. You are a product builder with deep AI fluency who will ship fast, write evals, configure agents, and sit with CX, Ops, Sales, Finance, and Clinical teams to find the workflows where AI actually moves the needle. Today, three engineers are building Eddy without a PM. You'll be the first, building the vision that the team is currently shipping without. You Will - Set & build the AI platform vision. Define what Headway’s agent harness should look like — from platform & data architecture to the connected ecosystem of agents, skills, and integrations that will sit on top of it. - Own Eddy's roadmap. Replace real-time feature triage with a first-principles product strategy. Build the user research pipeline, the eval infrastructure, and the prioritization framework that don't exist today. - Enable adoption across functions. Partner with functional leaders in CX, Ops, Sales, Finance, Clinical, and more to identify high-value AI workflows, build them, and turn non-adopters into power users. Who You Are - 8+ years of product experience, with at least 2 years shipping AI or AI-adjacent products. - You have personally built or configured AI agents, skills, or workflows — not just managed the engineers who did. You write prompts, design evals, and know where LLMs break. - Strong AI fluency: prompt engineering, context management, RAG, tool use, evals. You can hold your own with ML engineers and have real opinions on cost/latency/quality tradeoffs. - A platform thinker who can build for other builders, not only end users. - A cross-functional operator who translates AI capabilities into business value for non-technical stakeholders — and knows how to say "no" to low-value requests without losing the relationship. - Ruthlessly prioritized. You can hold a product roadmap, a platform roadmap, and an adoption strategy in your head at once and know which to pull on this week. - Above all, you believe AI is going to fundamentally change how knowledge workers operate — and you want to prove it at a company where the CEO agrees with you. We believe a team's strength is in its people, and we cannot achieve this mission without a team that reflects the diversity of this problem – across race, ethnicity, gender, sexuality, age, national origin, religion, family status, disability, military status, and experience. Headway is committed to the full inclusion of all qualified individuals. As part of this commitment, Headway will ensure that persons with disabilities are provided with reasonable accommodations. If reasonable accommodation is needed to participate in the job application or interview process, to perform essential job functions, and/or receive other benefits and privileges of employment, please inform the recruiter when they contact you to schedule your interview. Headway participates in E-Verify. To learn more, click here. A notice to Headway applicants: To protect yourself against phishing and recruitment fraud, please note that Headway only accepts applications through our official careers page at https://headway.co/careers. Headway will never refer you to external websites, ask for payment or personal information, or conduct interviews via messaging apps. All official communication will come from a @findheadway.com email address. If you are contacted by someone claiming to be from Headway via an unofficial channel, please do not share any information and report it as spam.

• Atuar de forma estratégica na nossa estrutura de Plataforma de Engenharia IA. • Referência técnica para a evolução de APIs, SDKs e serviços críticos de inteligência artificial na Vivo. • Definir a visão técnica e os padrões de desenvolvimento para APIs, SDKs e serviços de plataforma em IA. • Liderar a arquitetura de soluções complexas, garantindo que a orquestração de agentes e serviços de IA seja escalável e resiliente em nível corporativo. • Atuar como advisor técnico para lideranças e squads. • Evangelizar boas práticas de engenharia e cultura de qualidade em toda a diretoria de Dados & IA. • Desenhar e otimizar a governança técnica. • Identificar e mitigar riscos técnicos estruturais antes que impactem a operação ou o roadmap de produtos de IA.