Don't pay the learning cost twice: Persist what the human + AI team learned
Save the knowledge created along the way: why an approach failed, which architecture work better.
~/writing/2026-09-27-thirdpost.md
cesar@portfolio:~/home
ONLINE
[ok] portfolio kernel loaded[ok] static content mounted[ok] visitor session ready
IMG_001 // ordered-dither // 480x640
visitor@web:~$ ./whoami --verbose
# applied intelligence for real-world financial systems
Cesar Osorio
Applied AI | Financial Services | Credit Risk • Fraud Analytics • Open Banking
I build applied AI and machine learning solutions that turn financial data into useful decisions.
- location
- Toronto, Canada
- focus
- AI systems × financial data
- status
- building useful things
visitor@web:~$ cat ./about.md
01 / PROFILE
About me
I lead the development of applied AI solutions that transform financial data into better decisions.
My work combines machine learning, software engineering, and financial services. I focus on taking ideas from experimentation to reliable systems that solve real problems and deliver measurable value.
./applied-ai./credit-risk./fraud-analytics./open-banking
[view_full_profile --linkedin] ↗visitor@web:~/writing$ ls -lt --posts
02 / FIELD NOTES
Writing
3 files found
Could a financial institution build its own AI model that is faster, cheaper, and as good as ChatGPT or Claude ?
With the right training data, that specialized model can potentially get close to — and sometimes outperform — a much larger general-purpose model for that narrow task, while being significantly faster and cheaper to run.
~/writing/2026-09-17-secondpost.md
Are AI coding tools making deployment faster?
Not really. More code yes, but no more actual releases
~/writing/2026-09-03-firstpost.md
visitor@web:~$ ./connect --list
03 / CONNECT
Find me online
Interactive command line
Type help to see available commands.