Senior AI Engineer – Agents

MiraMace, Inc.; San Francisco (CA); Sur place; USD 150 000 - 210 000; Plein temps

Résumé du poste

Mira Mace is seeking an AI engineer to build the core of our product. You will own the agent layer, multi-step tasks, evaluation, retrieval, and the AI features used by internal users. You’ll work directly with the founders and ship production-grade systems built on modern LLMs.

Fluency with LLMs, strong software fundamentals, and comfort with ambiguity are essential. This role offers ambitious impact in a fast-growing healthcare AI startup based in San Francisco.

Qualifications

• You've built LLM-powered systems or agents and shipped them to production — not demos, but systems handling real usage by real people.
• Strong overall AI fluency. You can reason concretely about prompting, tool use, context management, retrieval, latency, cost, and failure modes.
• You've built evaluation into your work — designed metrics, captured signal from production, and used it to systematically improve quality.
• You've monitored and debugged AI systems in production, and you know what it takes to keep them reliable when real people depend on them.
• Solid software engineering fundamentals. You write code others can maintain, and you own what you ship.
• You're comfortable with ambiguity. The playbook doesn't exist yet, and you're excited to build it.

Responsabilités

• Build agents that do real work.
• Design and ship LLM-powered systems that take a multi-step task and see it through — deciding what to do, taking action across our systems and third parties, and knowing when to hand off to a human.
• Own quality and evaluation. Build the infrastructure that measures how well our AI performs, catches regressions before users feel them, and turns production signal into better prompts, retrieval, and models. Make quality a number we can move, not a vibe.
• Build the knowledge layer. Design the retrieval systems our agents reason over, and solve the harder half: keeping that knowledge accurate, versioned, and fresh as the underlying sources change.
• Build AI into the product. Ship copilots and assistive features that make our users meaningfully faster, with clear judgment about when the model should act and when it should ask.
• Close the loop on real usage. Deploy to production, watch how the systems behave with real users, and improve from what actually happens — in days, not quarters.
• Shape the technical roadmap. Work with the founders on what to build next. Bring a sharp sense of what current models can and can't do, and help us make smart bets on where the technology is heading.
• Lay the foundation for scale. Make architectural decisions that hold up as usage grows by orders of magnitude, and build with the next engineer in mind.

Connaissances

LLM systems

Agent development

AI fluency

Software engineering

Ambiguity tolerance

Outils

Python

TypeScript

FastAPI

Next.js

Postgres

Description du poste

ABOUT US

Mira Mace pairs Medicare beneficiaries with a dedicated healthcare advocate who navigates appointments, insurance, and care coordination on their behalf. Our customers get the support of caring nurses while AI agents handle the tedious backend work — all covered by Medicare. We've felt the pain ourselves — the endless back-and-forth with insurance, surprise bills, and the lack of clarity when you just need answers. Too many people fall through the cracks, and we're determined to change that. Today, 24/7 personalized health assistance is only available to the rich or extremely sick. Our vision is for everyone to be able to afford a health assistant who knows your health history deeply, navigates the healthcare system on your behalf, and propels you to become the healthiest version of yourself. Our founding team brings a mix of strong technical experience from companies like Google, Meta, Dropbox, and Amazon, along with serial startup experience ranging from early bootstrapped ventures to Series D scale-ups. We are backed by Foundation Capital, DefineVC and top Silicon Valley angel investors.

WHAT WE'RE LOOKING FOR

We're looking for an AI engineer to build the LLM-powered systems at the core of our product. Our product runs on agents. They decide what work needs to happen for each customer, do a lot of that work autonomously, check the quality of what got done, and assist the people on our team who handle the rest. You'll own that layer. Concretely, you'll build agentic systems that carry multi-step work from start to finish, the evaluation infrastructure that tells us whether they're any good, the retrieval layer that gives them accurate knowledge to reason over, and the AI features our internal users rely on every day. The surface is wide and still being defined, so there's a lot of room to shape what gets built. You'll work directly with the founders. We don't have a strong opinion about which corner of AI engineering you came from — agents, RAG, evals, fine-tuning, retrieval — it all translates. What matters is that you're genuinely fluent with modern LLMs, that you treat them as a backend you can build reliable systems on, and that you've shipped something real to real users.

RESPONSIBILITIES

• Build agents that do real work.
• Design and ship LLM-powered systems that take a multi-step task and see it through — deciding what to do, taking action across our systems and third parties, and knowing when to hand off to a human.
• Own quality and evaluation. Build the infrastructure that measures how well our AI performs, catches regressions before users feel them, and turns production signal into better prompts, retrieval, and models. Make quality a number we can move, not a vibe.
• Build the knowledge layer. Design the retrieval systems our agents reason over, and solve the harder half: keeping that knowledge accurate, versioned, and fresh as the underlying sources change.
• Build AI into the product. Ship copilots and assistive features that make our users meaningfully faster, with clear judgment about when the model should act and when it should ask.
• Close the loop on real usage. Deploy to production, watch how the systems behave with real users, and improve from what actually happens — in days, not quarters.
• Shape the technical roadmap. Work with the founders on what to build next. Bring a sharp sense of what current models can and can't do, and help us make smart bets on where the technology is heading.
• Lay the foundation for scale. Make architectural decisions that hold up as usage grows by orders of magnitude, and build with the next engineer in mind.

QUALIFICATIONS

• You've built LLM-powered systems or agents and shipped them to production — not demos, but systems handling real usage by real people.
• Strong overall AI fluency. You can reason concretely about prompting, tool use, context management, retrieval, latency, cost, and failure modes.
• You've built evaluation into your work — designed metrics, captured signal from production, and used it to systematically improve quality.
• You've monitored and debugged AI systems in production, and you know what it takes to keep them reliable when real people depend on them.
• Solid software engineering fundamentals. You write code others can maintain, and you own what you ship.
• You're comfortable with ambiguity. The playbook doesn't exist yet, and you're excited to build it.

NICE TO HAVE

• Depth in retrieval systems — RAG, vector databases, and keeping a knowledge corpus fresh over time.
• Experience with feedback-driven improvement loops for LLM systems (RLHF, reward modeling, or similar).
• Experience in healthcare or another regulated industry (HIPAA, PII/PHI handling).
• Python and TypeScript — our stack is FastAPI, Next.js, and Postgres.

WHO YOU ARE

• Proactive. You move quickly and take a forceful stand without being abrasive. You act without being told what to do and bring new ideas to the company.
• Analytically sharp. You structure and process qualitative or quantitative data and draw penetrating insights. You learn quickly and absorb new information with ease.
• High standards with attention to detail. You expect nothing short of the best from yourself and your team. You don't let important details slip through the cracks or derail a project.
• Passionate and open. You exhibit enthusiasm and a can-do attitude over your work. You solicit feedback often and react calmly to criticism or negative feedback.

OUR CULTURE

• Patient First. We start with the patient and work backwards. Every decision — what we build, who we partner with, how we operate — is filtered through one question: does this make the patient's life better?
• Sense of Urgency. Every day a patient goes unnavigated is a day someone needing help couldn't get the support they needed. We move fast, make decisions with conviction, and carry urgency toward the long-term vision: an AI nurse concierge in every patient's corner.
• Ownership. We see things through. We don't ship and walk away — we own outcomes, not just tasks. We act on behalf of the entire company, beyond just our own area of responsibility. We never say "that's not my job."
• Insist on the Highest Standards. We hold ourselves and our teams to nothing short of the best — in clinical quality, in operational execution, and how we show up for patients and partners. We raise the bar continuously.
• Question Everything, Unapologetically. We reason from first principles, not precedent. We challenge requirements regardless of who set them, dig until we reach the root of the problem, and resist the pull of "that's how it's always been done."

WHY JOIN US

Mission with massive impact. Every system you build puts a dedicated health advocate in someone's corner. We're building one of the largest AI-first companies in healthcare, and this is your chance to own core technical foundations early. A wide-open AI surface. Agents, evals, retrieval, copilots — and use cases we haven't scoped yet. This role has room to define its own scope, not a backlog to burn down. Learn fast, build fast. We believe in experimentation, measurement, and steady improvement. You'll ship in days, not quarters. Grow with us. You'll be part of the team that takes Mira Mace from early product to scale. The decisions you make now will define the company's technical DNA. Meaningful early equity. Competitive compensation and real ownership in what we're building.