Engineer — Agentic AI

Medfinder Concierge; Boston (MA); On-site; USD 100,000 - 150,000; Full time

Job summary

Medfinder Concierge is seeking an AI-native developer to build autonomous AI agents that function with little supervision. You will have direct access to the founders, helping shape product direction while focusing on automating repetitive tasks through innovative workflows.

The role emphasizes creating self-improving agents and requires a user-centric approach to product design. Enjoy working in an in-person environment with a lean team in Downtown Crossing, Boston.

Qualifications

• Experience in orchestration of multiple parallel AI agents.
• Ability to convert repetitive tasks into workflows.
• Strong understanding of self-improving agent design.

Responsibilities

• Build AI agents that require minimal supervision.
• Shape product direction and execute on improvements.
• Design and implement auto-research loops for self-improvement.

Skills

Building autonomous AI agents

Auto-research loops

Strong product judgment

Ability to work autonomously

Job description

At Medfinder, we build AI Agents that help patients get access to medications. We use AI agents (primarily voice AI) to automate manual processes in the prescription journey, and have already helped over 30,000 patients across the US get their prescriptions filled. Medfinder is the #1 medication locating service in the market, having partnered with hundreds of healthcare organizations and leading pharmaceutical manufacturers. We are growing extremely fast and lean, quadrupling our revenue and customer base in the last 12 months without expanding our team size.

We believe that everyone deserves access to the care they need, and are just at the start of our journey to make that a reality.

About the Role

This is not a traditional software development role. You are not building machines — you are building machines that build machines.

We want someone who is AI-native: not someone who uses Cursor or Copilot inside an IDE, but someone who thinks in terms of multiple parallel agents working semi or fully autonomously. You think in agents, not commits.

The gist: build autonomous agents that require minimal supervision. They should behave more like coworkers than tools, taking ownership of a problem, figuring out the next step, and coming back with results instead of questions.

Beyond building specific agents, we want someone who automates repetition. Anything you find yourself doing repeatedly in product development should be converted into an agentic workflow. Over time, most of the work becomes running, improving, and orchestrating those workflows.

The pattern we care most about is the auto-research loop: a self-improving agent that iterates against a metric, keeps what works, discards what doesn't, and gets better overnight without you. Designing these loops well is the main skill we're hiring for.

You need to know how to accurately provide context to the AI agents, encode tribal knowledge into AI-accessible documentation, design tools to safely work with production systems, and curate guardrails to prevent disasters and avoid AI slop. These aspects are crucial to ensure AI agents actually improve day-to-day productivity.

You'll be doing this at a company that already works this way. Our AI agents call pharmacies on behalf of patients every day, and this role takes the same approach to the rest of the product.

What You’ll Do

• Build autonomous AI agents that need little supervision and behave more like coworkers than tools.
• Work directly with the founders on what to build, not just how to build it. You'll shape product direction alongside us, then build the agents that execute on it. No middle managers, no PMs translating your work, no waiting for approval to ship.
• Convert anything you find yourself doing repeatedly in product development into an agentic workflow.
• Design auto-research loops for self-improvement. This is the pattern we care most about.

What We’re Looking For

• AI-native, not IDE-native. You don't just use Cursor or Copilot; you orchestrate multiple parallel agents working semi- or fully autonomously.
• You live on the frontier. You were the first one of your friends with an OpenClaw setup. You adopted agentic development before your peers. You see what's coming, and where the world is going, and you know you'll be at the front of the line.
• You’re excited about working in person with a lean team. We work in person in our office in Downtown Crossing in Boston with a strong in-person culture. We believe building in person is critical to success, and a whole lot more fun!
• You thrive in ambiguity, crave autonomy, and can't stand red‑tape. We have no middle managers, no PMs translating your work, no waiting for approval to ship.
• You own problems, have strong opinions, and figure it out. You don't wait to be told what to do; you find the next thing worth doing and ship it.
• You think like a user. You have taste and product judgment. When an agent returns three possible designs, you know which one is right and why.
• You can write the spec. Knowing what to ask the AI to do is half the job. You can scope work, write clear briefs, and reason about tradeoffs the way a PM would.

Bonus Points

• You're a former founder, or a future one.

How to apply

Have your agent apply on your behalf using our MCP server: hatch-one.vercel.app/api/mcp