Understand the problem before choosing the tool
I do not begin with a framework or model. I begin with the user, constraints, cost of failure, and evidence of success, then choose the technology.
ABOUT / BUILDING IN PUBLIC
I connect operational problems, system design, and AI-augmented delivery.
This is more than another version of my CV. It is a verifiable record of how I start with an operational or user problem, define system boundaries, and use AI to accelerate prototyping, testing, and delivery while stating ownership and limitations clearly.
Problem framingProduct judgementVerified delivery
Three connected capabilities
My value is not pretending one person replaces a full engineering team. It is making the operational problem clear, defining system boundaries, and turning the proposed solution into evidence that can be tested.
01 / Operations and architecture
Make risk, exceptions, and ownership visible inside the system
Connect SDKs, strategy lifecycle, controls, confirmed state, and observability through an operational lens.
Inspect the evidence →02 / Connectivity and product judgementTurn fragmented data into a stable decision interface
Define reading order, data boundaries, failure states, and acceptance criteria before completing the usable product.
Inspect the evidence →03 / AI-augmented deliveryAccelerate execution without outsourcing judgement or accountability
AI supports research, implementation, and QA; I own the proposition, trade-offs, verification, publishing boundary, and final quality.
Inspect the evidence →AI expands execution; judgement remains human
AI agents help me research, code, compare options, and inspect results. I remain responsible for framing, trade-offs, testing, final quality, and what becomes public.
Evidence before polish, privacy by default
Working products, tests, and honest limitations matter more than polished claims. New material starts private and becomes public only after redaction and human review.
My build loop
AI can accelerate each step, but it cannot replace clear ownership. Every public project passes through problem framing, implementation, verification, and a safety review.
01Frame
Define the user, problem, constraints, and success criteria.
02Explore
Use agents to research options, risks, and viable paths.
03Build
Complete the smallest workflow that is genuinely usable.
04Verify
Test behaviour, data, responsive UI, and public claims.
05Publish
Redact, visualise, and preserve honest limitations.
Public does not mean publishing everything
This portfolio demonstrates capability and judgement without using employer, client, household, or account data as proof.
Safe to publish
- Original tools, usable demos, and product design
- Sanitised system architecture and engineering trade-offs
- Academic research, public data, and reproducible charts
- Edited and human-reviewed learning records
Stays private
- Real employer, client, and internal workflow details
- API keys, credentials, accounts, and infrastructure settings
- Household, identity, and real financial data
- Strategy parameters, positions, orders, and live results
CONTACT
Let’s make the idea clearer and get it working.
I am happy to compare notes on public AI projects, agent workflows, trading and data system design, and the work of turning prototypes into useful products.