Building decision intelligence that shows its work.
I am developing Azumie through a current small-business client engagement, using real operational questions to learn what trustworthy decision support must explain, preserve, and help people do next.
NOWbuilding from real decisions
ROLE
Founder · Product and research
WHEN
2026–present
WITH
Current small-business client engagement
Azumie began as a broad interest in intelligence infrastructure. Working with a real small business is forcing that ambition into a sharper question: what information does someone need before they can confidently act on a recommendation?
The current work concentrates on operational decisions and the data behind them. The machine-learning layer is treated as a research direction to earn through clean evidence and evaluation, not a feature label added before the underlying workflow is trustworthy.
Azumie is being shaped as a decision workspace rather than a generic chatbot: recommendations remain connected to their evidence, calculations, assumptions, risks, and next actions.
03
THE PROCESS01 02 03
How I moved through the work.
Observe the decision
Start with a repeated operational choice and map the information, constraints, judgment, and uncertainty involved before designing an interface.
Preserve the evidence
Keep recommendations connected to the sources, calculations, assumptions, and risks a decision-maker would need to verify them.
Evaluate the intelligence
Introduce forecasting or recommendation methods only where they can be compared against clear baselines and useful outcomes.
04
KEY DECISIONS
The choices that shaped the result.
- Begin with operational decisions instead of serving every industry and decision type at once
- Design a workspace for evidence and action rather than another chat interface
- Keep human review visible wherever business context or uncertainty matters
- Use client work to generate research questions without presenting unverified ML capabilities as finished
WHAT IT DELIVERED
Defined a focused starting point in operational decision intelligence
Grounded product discovery in a real client environment
Connected product work to applied ML and trustworthy-AI research questions
MY SCOPE
Founder research
Product strategy
Decision-system design
Applied ML
Azumie is teaching me that trustworthy intelligence begins before the model. It begins with understanding the decision, preserving the evidence, and making uncertainty useful enough that a person can choose what to do next.