We build AI you can rely on: answers that cite their sources, an assistant that stays on your own devices, and our own models, from LLMs to live trading systems.
Products
AI should do real work for people, show how it got there, and leave them in control.
Legal and public-procurement intelligence for Panama.
A personal AI assistant that lives on your own devices.
Research
We train our own language models. Separately, we build neural-network trading systems end to end, from market data and retraining to execution, and run them in live markets, the hardest test of speed and consistency. What we learn shapes every product.
Training and fine-tuning LLMs in-house, for legal research and assistants that run on your own devices.
Transformers, recurrent and convolutional networks and hybrids, compared head to head on live market data, retrained on rolling windows and selected only on data they haven't seen.
Market data to trading decision, typically in under 5 milliseconds, with every step profiled in production rather than assumed from a backtest.
Live results reconciled against simulation, and code changes replayed on recorded market data before they ship, with drift monitoring and automatic safeguards.
Approach
A small company can hold a high bar when every part of the product has the same owner.
Answers come with their sources and are checked against the latest version of the facts. When something can't be verified, the product says so. We hold our live trading systems to the same standard.
Personal data stays as close to its owner as possible. ScarlettOS runs on your own hardware and keeps private things for your voice only.
Bilingual from day one and grounded in local institutions, with products designed to grow beyond one market.
Founder
“I've been building since I was twelve, starting with robots. I studied AI because I believed it would one day be the most powerful tool in any field, once the models were good enough. That day has arrived, and now I want to put what I've built in people's hands.”
I build everything myself, from research and data pipelines to the apps people use. Before going independent I was a Generative AI Engineer at Amazon in Seattle. A search engine I prototyped on my own grew into a dedicated product team, and I built coding agents that migrate large systems between programming languages.
Taliaferro Technologies is where years of experiments become tools for everyone. My first AI legal assistant for Panamanian law, built in 2023, grew into ForsetiLex. ScarlettOS is the assistant I always wanted: one that lives on your own devices and speaks both of the languages my household speaks.
— Sam
ForsetiLex, ScarlettOS, and research training our own language models and live neural-network trading systems.
AI search, coding agents for large-scale code migration, and a legacy promotions system replaced at scale.
Built and ran an AI legal assistant with a self-updating database of Panamanian law.
Automated anti-money-laundering news screening, and cut property legal descriptions from months to minutes.
Designed the first vertical hydroponic towers used in Panama.
Started building robots, and never stopped building.
Questions about ForsetiLex or ScarlettOS, partnerships, or AI work in Panama: reach out directly. You'll hear back from the founder.