Staff Engineer, AI Developer Platform

Staff Engineer, AI Developer Platform

Remote (US) · Full-time

Company Overview

b.well is solving healthcare’s fragmentation problem with our FHIR-based health data management platform. The platform connects data from EHRs, wearables, portals, and other sources, while our intelligence engine personalizes the consumer experience. By simplifying the complex healthcare ecosystem, we make it easy and convenient for consumers to engage and take action—whether it’s scheduling care, setting reminders, accessing health data, and more. For our clients, this means better health outcomes, operational efficiency, and stronger consumer engagement.

Position Overview

AI is changing how software gets built, and we want b.well’s engineering organization to be among the teams that do it well: faster delivery, without trading away quality, security, or reliability. We’re hiring a Staff Engineer to build the platform that makes that possible.

You’ll own the internal AI developer platform that every engineering team builds on. That means agentic workflows embedded in our software development lifecycle, the context and cost controls that keep them efficient, the guardrails that keep them safe, and the evals and metrics that tell us whether they’re actually working. Your customers are b.well’s engineers, and your job is to make the AI-native path the easiest path.

This is a hands-on role. You design these systems, you build them, and you help operate them in production. You work horizontally across teams rather than inside a single product squad, so the workflows and standards you create show up everywhere. The hard problems here are less about getting an agent to work once and more about making agents dependable, measurable, affordable, and safe at the scale of a whole engineering organization.

This role is focused on engineering productivity and quality. AI features in our customer-facing products are owned by other teams, though you’ll partner with them on shared practices.

What you’ll do

• Build and run the internal AI developer platform as a product: golden paths, self-service tooling, documentation, and adoption metrics, with engineers as your customers.
• Design and ship orchestrated agentic workflows across the SDLC, such as code review, test generation, migrations, dependency upgrades, and incident triage, and decide where a deterministic workflow beats an agent.
• Build and govern the tools agents use, including MCP servers, tool schemas, and reusable skills, so they’re reliable across teams.
• Engineer context deliberately: retrieval and codebase indexing, context compaction, and keeping long-running sessions focused and accurate.
• Manage token spend as a first-class concern through prompt caching, model routing and tiering, budgets, and per-team cost visibility.
• Implement guardrails for agents with real permissions: least-privilege access, sandboxed execution, approval gates for destructive actions, prompt injection defenses, and audit trails.
• Keep sensitive data where it belongs. In a healthcare company, logs, tickets, and incident data can contain protected health information, and agent workflows need to respect those boundaries by design.
• Build evaluation into the platform: task-based eval suites drawn from our own codebase, regression checks when prompts, models, or tools change, and production monitoring that catches quality drift.
• Define and report productivity and quality metrics that leadership trusts, connecting AI usage to delivery speed, change failure rate, rework, and defect escape rather than vanity numbers.
• Evaluate commercial AI developer tools against what we should build ourselves, and integrate rather than reinvent where it makes sense.
• Set technical direction for AI-native engineering through RFCs and roadmaps, and drive adoption across teams through enablement, office hours, and internal champions.
• Partner with Security, Enterprise Architecture, and engineering leadership on standards, policy, and rollout.

What we’re looking for

• Deep platform engineering experience. You’ve built and operated internal platforms or developer tooling that other teams depended on, with real adoption, SLOs, and on-call ownership.
• You’ve shipped agentic workflows that other engineers actually used in production, and you can talk candidly about what failed, what you changed, and what you abandoned.
• An evaluation mindset. You measured whether your AI workflows worked and changed course based on the data. You don’t need to have built a mature eval platform, but you need to know how you’d tell good from bad.
• Strong security instincts for systems that act on their own: permissions, sandboxing, approval gates, and awareness of how agents get misled or misused.
• Staff-level influence. You’ve driven adoption across teams without formal authority and written technical strategy that others followed.
• Solid foundations in CI/CD, cloud infrastructure, and observability. Our environment includes AWS, Kubernetes, infrastructure-as-code, OpenTelemetry-based observability, and a polyglot backend across Python, TypeScript/Node.js, and Java.
• Fluency with AI coding tools like Claude in your own daily work.

Nice to have

• Built eval harnesses or regression suites for LLM-based workflows, including calibrating LLM-as-judge scoring against human review.
• Hands-on LLM cost optimization: caching, routing, budgets, and chargeback or showback models.
• Designed MCP servers or tool ecosystems used by multiple teams.
• Measured engineering productivity rigorously using frameworks like DORA, SPACE, or DX Core 4, with baselines and cohorts.
• Agent observability and tracing experience.
• Experience in healthcare or another regulated industry such as fintech or govtech. You’ll partner with Security on HIPAA requirements, but we expect you to design with sensitive-data boundaries in mind from day one. If you haven’t worked in healthcare yet, that’s fine; it’s learnable, and we’ll help you get there.

What success looks like

• Engineers across b.well use AI-native workflows by default because they’re the fastest, most dependable way to get work done, not because they’re mandated.
• We can show, with data, that AI is improving delivery speed without increasing change failure rate, rework, or escaped defects.
• Agent workflows are covered by evals, and we catch quality regressions from model, prompt, or tool changes before engineers do.
• Token spend is visible, attributed, and managed, and we can explain the return on it.
• Guardrails are strong enough that Security trusts agents with meaningful permissions, and no sensitive data leaves approved boundaries.
• Other engineers have grown into AI-native practices because of your enablement and mentorship.

Compensation & benefits

The target salary range for this position is $225,000 - $275,000 and is part of a competitive total rewards package including stock options, benefits, and incentive pay for eligible roles. Individual pay may vary from the target range and is determined by a number of factors including experience, location, internal pay equity, and other relevant business considerations. We review all employee pay and compensation programs annually at minimum to ensure competitive and fair pay.

Data shows that women, people of color, and other underrepresented groups may be less likely to apply for jobs unless they believe they are a perfect match. But b.well holds diversity amongst its key values, and we have a strong commitment to building our workforce and products through that lens. You don’t have to check every box in this job description to be a great fit for the role!

If you’re excited about this position and the prospect of working for b.well, please apply. If it turns out this role isn’t for you, there may be other openings that could align with your experience and expertise!

We are an equal opportunity employer. We do not discriminate based on race, ethnicity, color, ancestry, national origin, religion, sex, sexual orientation, gender identity, age, disability, veteran, genetic information, marital status or any other legally protected status.