ai-lawtech

Richard Careaga · AI trainer · Legal Engineer

AI that works the way lawyers and recruiters need it to.

I train frontier models for talent-as-a-service bureaus to prompt and verify AI output, and I apply AI tools to legal engineering work, with the scoring discipline to know when the answer is actually usable, including preparing golden responses.

See examples ↓ Connect on LinkedIn
  • Senior Vice President/Associate General Counsel in charge of a major bank's mortgage-backed securities business
  • AI trainer on professional-grade legal and financial tasks
  • Rubric designer for accuracy, autonomy and verification

What I do

Two ways I put AI to work.

AI training for TaaS shops

Recruiters placing legal and professional talent need to see that I know what good AI output looks like and how to tell when it isn't.

  • Writing prompts that create ambiguity to break the model
  • Writing prompts that avoid ambiguity to exercise the model
  • Scoring output against a clear rubric
  • Spotting anachronisms, unsupported claims and false confidence

AI for legal engineering work

Drafting, research and analysis workflows built around the way legal work is actually checked and signed off.

  • Issue-spotting across tax, accounting, contract and disclosure questions
  • Memoranda written from the expert, not to the expert
  • Verification steps before anything is called finished
  • Rapid prototyping of new workflows

Examples

Look at the work first.

Three examples: a reusable golden brief, a quotation verifier, and an AI-assisted privilege review.

Golden response

Healy v. Milliman, briefed for reuse

A 21-page, non-adversarial reference entry on a 2026 Ninth Circuit decision about class-wide standing at summary judgment, with sources reviewed and rule text checked.

Read the brief → Verification

Catching the quote that is almost right

A deterministic checker against the full United States Reports sorts quotations into exact, typo, altered and nonexistent, in under a second.

See the checker → Legal engineering

Privilege log from 249,068 emails

A case study in directing an AI agent through the Enron corpus: the human review queue fell to 8,384 priority messages (3.4%) in a single working session.

Read the case study →

Seen enough? Let's talk.

I am an AI evaluator, a former Associate General Counsel and Administrative Law Judge, and an AI and data science researcher. See my portfolio.

Recruiters hiring for remote, part-time AI training or legal engineering roles: message me on LinkedIn or write to richardcareaga@protonmail.com. Tell me the specifics, and I'll tell you whether I'm the right reviewer.

Contact me on LinkedIn →