Founding Engineer

Flux Cap. is building the AI-native operating system for high-stakes litigation. Premium legal work hasn't fundamentally changed in decades, giving massive advantages to the largest firms and corporations. We're rebuilding the infrastructure from scratch to help David take on Goliath in billion-dollar legal battles.

Founded by the former heads of the Department of Justice's Antitrust Division, backed by $6M+ in seed funding from YCombinator, Bloomberg Beta and others.

You would be one of the first engineering hires, working directly with the CTO and founders. Full-stack IC work: React/Next.js frontend, Typescript backend, Postgres with pgvector for the semantic layer.

You'll sit with lawyers who ran the DOJ's biggest antitrust cases and translate how they reason into systems that can do it at scale. Legal reasoning isn't a search problem. A single document might strengthen one claim, undermine another, and be irrelevant to a third depending on which legal theory you're pursuing.

• Design the data layer that connects claims to elements to facts to evidence, structured enough for agents to reason over without hallucinating citations or skipping steps

• Build entity extraction workflows that pull structured claims, actors, and timelines from legal files

• Create agent workflows that surface gaps in legal arguments before a complaint gets filed

• Build evaluation frameworks where attorney corrections flow back into the system and sharpen future retrieval across every case

Requirements

• 3+ years shipping production software

• Strong with React, TypeScript, Next.js

• Comfortable with SQL/Postgres (bonus: vector search, graph-like queries)

• Based in NYC or willing to relocate. Hybrid NYC role.
  

• Value autonomy over process

• Curious about the domain

Experience with LLMs, RAG, or evaluation frameworks is a plus. Interest in learning it is required.

Compensation

$175K-$210K based on experience, plus meaningful equity.

Benefits: Health, dental, and vision insurance. 401(k)

Interview Process

1. 45-min technical interview with CTO

2. Paid work trial (1 days onsite in NYC; we cover travel, food, housing). You will work on a real problem with the team

3. Founder conversation

4. References

Consider This Problem

The first step in any legal matter is the drafting and filing of the complaint. The complaint makes claims of harm incurred by a company. Each claim has elements: the specific things that must be proven for the claim to succeed. Each element requires facts, and each fact needs supporting evidence.

For example, a price-fixing claim might require proving: (1) an agreement existed between competitors, (2) the agreement affected prices, and (3) the plaintiff was harmed. Element 1 might be supported by facts like "executives from Company A and Company B met at a trade conference in March 2023" and "prices increased uniformly across both companies within 30 days."

Now imagine you have:

• 10,000 documents (emails, meeting notes, SEC filings, earnings calls)

• A complaint draft with 5 claims, 15 elements total

• Each element tagged with the facts that support it

The questions:

1. How would you represent this structure in a database such that a lawyer can ask "which elements are weakly supported?" and get an instant answer?

2. An attorney reviews the system's work and says "Document X doesn't actually support Claim 1; it supports Claim 5." How do you capture that correction in a way that improves future suggestions?

   • What if one part of Document X supports Claim 2, but another part undermines Claim 2?

1. Six months later, discovery produces 500,000 new documents. How does your system identify which new documents are relevant to which elements, without re-processing everything from scratch?

This is a simplified version of what we are building. If it sounds interesting, we would like to talk.

Our team: DOJ prosecutors who've litigated billion-dollar cases, Palantir engineers who've built mission-critical systems, and White House advisors who've operated at the highest levels.
You must be authorized to work in the US. We are not currently able to sponsor visas.