Learning systems since 1993. Now AI applications for professional firms.
CleanAim® was built by a founder who has been building capture-pair-learn systems since 1993. Today we build and maintain AI applications for professional firms with CleanAim® Studio.
THE FOUNDER
From enterprise learning to AI applications: the same methodology, applied to the work of professional firms.
Tom Killi has been building systems that capture decisions, pair them with outcomes, and enable compound improvement since 1993. The methodology has evolved across four companies, but the core architecture has remained the same.
Where x.hlp paired decisions by people with learning results over days, CleanAim® pairs AI decisions with outcomes in milliseconds. The principle is identical. The speed is different.
The patent, published in 2001, has been cited by patents from Microsoft, Apple, Samsung and IBM, and the technology ended up inside SAP. The same methodology now underlies the AI applications we build for professional firms.
THE TECHNOLOGY ASSET
Self-funded R&D. The platform, patents, and methodology were built with conviction, not committees.
CleanAim® is a fully self-funded technology asset with clear IP ownership and zero external obligations.
1.8M+ lines of production code, across four generations of CleanAim®. 7 LLM providers supported. 93.3% cross-model transfer efficiency (internal testing across Claude, GPT, Gemini and Grok; 250 tasks).
THE DISCOVERY
We pointed our own verification system at our own codebase
Every test passed. CI was green. Health checks said HEALTHY.
Then we ran it. It found 4 silently broken data pipelines, 17,952 calibration records that nothing had computed, and 131 mutations where there should have been eleven thousand.
None of this showed up in tests, monitoring, or code review. That discovery became how we verify.
THE TECHNOLOGY
Numbers that matter
Built, measured, and verified. Not projected.
Lines of production code, four generations
LLM providers supported
Cross-model transfer efficiency (internal testing)
FOUNDER & CEO
Tom Killi
Founder & CEO
Tom Killi has built systems that learn from the prediction-outcome loop since 1993.
In 1996, he founded x.hlp Technologies, which developed capture-pair-learn methodology for enterprise training. The company served 30+ Fortune 500 clients including JP Morgan and Shell, plus McKinsey and NATO, and was acquired in 2004. That technology now operates inside SAP.
After building grid analytics and prediction-outcome systems in the energy sector from 2009 to 2019, he founded CleanAim® in 2020 to apply the same methodology to AI. Today CleanAim® Studio builds and maintains AI applications for professional firms.
In 2026 he published Silent Wiring: The Hidden Failure Mode in AI-Generated Code. It documents what he found when he pointed behavioral verification at his own codebase, and 21 independent incidents like it at Google, Uber, Cloudflare, Knight Capital and others. More about the book at silentwiring.com.
More about Tom, and everything he has built, at tomkilli.com.
Learning Systems Since
Patent, cited by Microsoft, Apple, Samsung, IBM
Fortune 500 Clients
THE MISSION
The AI prepares. Your people decide and sign.
We build AI applications that do the recurring work behind a firm's deliverables the way the firm does it, and we keep them current. Every figure can be checked against its source, and every decision and signature stays with the firm's people.
APPROACH
How we build
Four architectural principles guide everything we create.
Constitutional
Systems that enforce their own rules. Not guidelines that can be ignored, but architectural constraints that cannot be bypassed.
Compound
Continuous improvement, not one-time optimization. Systems that get better every day through the capture-pair-learn loop.
Portable
Intelligence that survives provider switches. 93.3% cross-model transfer (internal testing) means no vendor lock-in.
Honest
Systems that admit when they don't know. A 98/100 score with explanation is more trustworthy than 100/100 that papers over gaps.
COMPANY
Company facts
Tell us which piece of work you'd automate first.
A person reads every enquiry and replies.
Talk to us