Software Engineer (All Levels)

We’re fixing one of the most broken and crucial parts of plaintiff litigation: getting complete, accurate medical records. Today, that takes 30 to 50 days of faxing, calling, and portal-chasing across thousands of providers, each running its own back office.

We deliver records in ~12 days on average. To do this, we're building an agentic layer that operates in this messy, real-world environment. Every retrieval teaches our system how to access another corner of American healthcare, building a provider data layer you can't scrape or buy. All while putting us at the most upstream point of a $20B+ market where everything downstream depends on us.

This problem extends far beyond the plaintiff legal space. It's also pervasive in healthcare, where our CEO Alvaro first encountered it while running a home health agency, as well as in disability services, life sciences/clinical trials, and the insurance industry. All of these verticals face this same challenge.

“Codes isn’t just participating in the workflow; it’s capturing the moment where the most important data is created, positioning itself to expand from ingestion into orchestration and ultimately own the entire stack.” - Amplify Partners, Boring is Sexy

We've gone from $0 to a $3M+ run rate in a year and grown 2.5x quarter over quarter for four straight quarters. We've raised $17M from Amplify, General Catalyst, Haystack, and Y Combinator. We serve 130+ law firms across plaintiff litigation and mass tort, including the largest plaintiff firm in the country.

The core challenges are data complexity and reliability under extreme variability.

You can think about it as reverse-engineering the back office of every healthcare facility in the U.S., then rebuilding it as a system that continuously learns from outcomes and improves across thousands of edge cases.

That includes:

• Building agentic systems that run end to end across fax, email, web portals, and physical mail, where each document type triggers a different workflow.

• Processing records that average 300 pages, sometimes hit 30,000, and run roughly 70% duplicate, and still surfacing what matters to a case.

• Matching patients and facilities across messy provider networks, where a single system like NYU Langone is 340+ care locations with no clean way to match them.

We're a lean team where everyone works on everything. The work splits loosely across automating the retrieval pipeline, the customer-facing product law firms use to request and view records, and the applied AI threaded through both.

You'll move between them as our goals shift, ship customer-facing product one month and go deep on the pipeline the next, and stay close to both the customer and the data.

This role is open at all levels. You might be a fit if you:

• Have shipped features start to finish in production and want to own outcomes.

• Go deep on the domain: the data systems and the workflows our agents run.

• You’ve explored LLMs and are excited about turning them into reliable systems.

• Want to talk to users directly and ship off their feedback.

• Are excited to apply real engineering rigor to a messy, economically valuable problem.

You'd work directly with the founders and a small, high-bar engineering team.

Fully in person in Dumbo, Brooklyn.

TypeScript and React on the frontend. Python and FastAPI on the backend.