Director of AI Research · Truth Computing
Johnathan Fierro
AI research engineer. He builds the tests that show whether an AI system's answers can be trusted, and the software those systems run on. At Truth Computing he works on Clientlyy, client communication for personal injury law firms.
The question he works on
How do you know when an AI system has earned your trust?
He answers it with measurements: scoring Clientlyy against labels that lawyers supply, and testing whether the model that checks its citations gives the same verdict twice.
Core capabilities
01
AI evaluation and reliability
Expert-labeled test sets and a pipeline that scores the model against them. He built Clientlyy's, with lawyers supplying the labels, and measured whether its citation checker holds its verdict.
02
Systems that learn from corrections
Software that learns a firm's preferences from the edits its people make, and lists every rule it learned where a person can switch it off.
03
Statistical modeling and simulation
Forecasts made by running the same situation thousands of times and counting how often each outcome happens. His sports model works this way.
04
Products and infrastructure
React and Flutter apps, Firebase back ends, live dashboards, and one-command setup for each new Clientlyy firm.
Selected work
-
01
2026 · PresentClientlyy
Client communication for personal injury law firms
Built the model-evaluation pipeline for Clientlyy, where an attorney labels a triage corpus and the system scores Clientlyy's classifier against those labels, and measured the grounding verifier for reliability, position bias, and framing effects. Built "learn from edits," which infers how a firm writes from what its attorneys change and surfaces what it learned, and one-command Fly.io provisioning so a new firm can be stood up without a manual setup pass.
-
02
Jan – Apr 2026Ntry
Phone-based door access for college dorms
Built to replace dorm keys. He led development with a team of three, designed the hardware in Autodesk Fusion 360, and built the companion app in Flutter for iOS and Android with live data through Firebase. Students get in by Bluetooth proximity or a QR scan.
Unlocking a door from the Ntry app. The device is an animation built on the enclosure design; the phone is the real app with demo data.
-
03
May 2024 – PresentSports Prediction Algorithm
Independent sports forecasting model
Runs each matchup thousands of times on live game data and counts how often each side wins, and reached 70% accuracy in his own testing. A Claude-powered assistant answers questions about the numbers in plain language, inside a React app where you can watch the simulations run.
-
04
March 2026Polymarket Wallet Tracker
Live prediction-market dashboard
Tracks positions, trades, and profit and loss across wallets as they happen, with alerts and trader rankings. Data pipelines pull from market APIs to keep wallet and market data current.
Record
- Present Director of AI Research Truth Computing. Evaluation, reliability testing, and infrastructure for Clientlyy.
- In progress B.S. Computer Science California Baptist University. Focus on machine learning and AI.
- 2019 National Champion CodeMonkey CodeRush National Championship, grades 6–8 division. Led his middle-school team to the title.
- Since 5th grade First line of code Scratch.
What Truth Computing stands for
Trust
Every claim we make, we can prove.
Discipline
We do fewer things, and we finish them.
Creativity
We find the answer no one else was looking for.
Execution
We own the outcome, not the recommendation.
Exuberance
We love this work, and it shows.
Get in touch
Have work that
has to be right?
Send a note or put time on his calendar, and bring the problem you cannot afford to get wrong. If it is a fit, he will tell you exactly how he would rebuild it.