Full-Stack AI Engineer

Full-Stack AI Engineer — Liza

About Liza

Liza Health is a Cambridge-based startup building an AI-enabled platform for modern primary care. Our bet is simple: primary care is the foundation of a healthy US health system, and AI finally makes it possible to rebuild and scale primary care from first principles instead of patching what's broken.

Liza was co-founded by Dr. Rushika Fernandopulle, a practicing physician who previously co-founded Iora Health, the relationship-based primary care company later acquired by One Medical and then Amazon, serial tech entrepreneur and investor Andy Palmer and healthcare serial entrepreneur Duncan Reece. We bring decades of experience redesigning care delivery together with a team that builds with frontier AI every day.

We work alongside primary care practices to change how care is delivered: more time for relationships, less time on administrative burden, and patients who get the right care at the right time. If you want your engineering work to make care better for real patients and clinicians, this is the place.

The role

We're hiring AI Engineers who will design, build and ship the AI systems at the core of Liza's care model. You'll own problems end to end: sitting with clinicians and care teams to build real-world healthcare applications for patients, clinicians and clinical operators.

This is a generalist role by design. On any given week you might write a prompt and its eval suite, wire an agent into an EHR via FHIR, build a frontend feature for care coordinators, and develop healthcare data pipelines. We value curiosity, learning fast and shipping well. We don't care much about titles or hierarchy.

The AI engineering bar here is the one set at the top AI labs: we A/B-test agent changes in a harness, LLM judges score replies in production, and we review production traces every week. The difference is where the work lands: in the hands of patients and primary care teams across the US, helping real people live healthier lives while reducing overall healthcare costs.

What you'll do

• Build production AI applications and agentic systems that support patients, clinicians and care teams across scheduling, intake, documentation, care coordination, outreach and clinical decision support.

• Partner directly with patients, physicians, health coaches and operations staff to turn messy, real-world workflows into clear technical requirements and measurable outcomes.

• Work closely with our AI-native design team to create experiences that are simple, efficient and effective for all of our users.

• Design, test and iterate on prompts, context and tool definitions that shape how our models behave, and keep them versioned, reviewed and safe to change.

• Define and run evaluations: build gold-standard datasets with clinicians, write rubrics and automated graders, read transcripts for what the metrics miss, and set launch criteria before anything reaches a patient.

• Integrate AI with our healthcare stack: our FHIR backend (Medplum), practice EHRs, HIE aggregators such as Metriport, HL7 interfaces, claims and scheduling systems, and our data platform.

• Build with PHI, HIPAA, privacy, security and auditability in mind, including human-review workflows and clear escalation paths for clinical risk.

• Monitor systems in production and use error analysis, observability and user feedback to rapidly improve reliability, cost, latency and model selection.

• Turn what we learn into reusable building blocks: shared eval harnesses, agent patterns, integration libraries and internal tools that make the whole team faster.

Who we're looking for

We want exceptional builders who believe in our mission and want to have a great time pursuing it. You'll fit right in if you are:

• A mission-driven believer. You think US healthcare can be dramatically better, that primary care is where that change starts, and that AI is the lever that makes it possible now. You want your work measured in better care, not just shipped features.

• A generalist. You're comfortable across the stack and across problem types, from backend services and data pipelines to frontends, prompts and evals. You pick up new tools, domains and languages quickly, and you go wherever the problem is.

• An exceptional builder. You take pride in shipping things people actually use at large scale and you have a track record of doing it. You move from idea to working prototype fast, then have the discipline to make it reliable. Startup experience is deeply appreciated at Liza.

• A strong teammate. You do your best work on small, cross-functional teams with clinicians, designers, operators and other engineers. You communicate clearly, share context generously, give and take direct feedback, and care more about the team winning than about who gets credit.

• Someone who has fun. You bring curiosity, humor and energy to hard problems. You like celebrating wins, laughing at the inevitable chaos of a startup, and building something meaningful with people you enjoy.

• Someone comfortable with ambiguity. You can frame an unclear problem, make good calls with incomplete information, and own the outcome.

No prior healthcare experience or clinical credentials are required. We'll teach you the domain; we're hiring for curiosity, judgment, craft and heart.

Essential AI skills

Strong candidates will be deep in several and conversant in all.

• Production LLM application engineering. 5+ years of software engineering (our stack is TypeScript end to end), with hands-on experience shipping LLM-powered products to real users at scale.

• Prompt and context engineering. Designing, testing and optimizing system prompts, tool descriptions and retrieved context, and knowing each frontier model's strengths and failure modes; treating prompts as versioned, reviewed production code.

• Agents and tool use. Building agentic workflows with function calling, multi-step planning, memory and state management, and protocols like MCP, including multi-agent systems with human oversight built in.

• Evaluation and error analysis. Building datasets and rubrics with domain experts, LLM-as-a-judge and automated graders, human rating workflows and regression detection, then reading transcripts, finding failure patterns and iterating until quality measurably improves. Evals are how we know a system is safe enough for patients.

• Retrieval and knowledge systems. RAG pipelines, embeddings, structured knowledge and memory that let models reason reliably over large, messy, heterogeneous data such as longitudinal patient records.

• Full-stack and systems breadth. APIs, data infrastructure, web and mobile frontends, cloud deployment, CI/CD, containers and production observability. You can take a system from prototype to reliable production on your own.

• AI-native development. Fluency with coding agents and AI-assisted workflows to multiply your own output.

Like to have

• Experience with EHR and interoperability technologies such as Epic, Oracle Health (Cerner), athenahealth, HL7, FHIR, SMART, health information exchanges and data aggregators such as Metriport, Zus, Gorilla Health, Redox.

• Exposure to primary care, value-based care, care coordination or patient access workflows.

• Experience fine-tuning or post-training models, or building data programs for supervised fine-tuning and reinforcement learning.

• Voice or multimodal AI experience, such as patient-facing phone agents or ambient documentation.

• Prior work as a forward-deployed engineer, technical founder or early engineer at a startup.

• Open-source contributions, publications or side projects that show your craft.

How we work and what we offer

• Small, cross-functional teams. Engineers, clinicians, designers and operators work side by side. Every person has real ownership and direct impact on patients.

• One codebase, shared by everyone. Our APIs, web and mobile apps, infrastructure and agents live in one TypeScript monorepo, and every engineer ships across all of it, with coding agents as everyday tools.

• Speed with rigor. We prototype fast, test with real users early, and hold a high bar for safety and quality before anything reaches care.

• Fun is part of the job. We take the mission seriously, not ourselves. Expect team lunches, hack days, celebrating wins together, and colleagues you'll genuinely enjoy.

• Location: Cambridge, MA, with in-person collaboration. Hybrid encouraged - most people are in the office 2-3 days/week but we are flexible especially after ramp up.

• Compensation: Competitive salary and meaningful equity, plus comprehensive health, dental and vision benefits.

Liza is an equal opportunity employer. We value diverse perspectives and backgrounds, which make us better at building care for everyone.