Approach
The decisions behind the software.
I like finding the simpler way through a difficult problem, whether that's the product plan, the architecture or how the team works.
Product and technical strategy
The product plan belongs in the technical discussion.
A technology choice sits alongside decisions about product scope, budget and who will deliver the work. I want a technical roadmap to make those connections clear.
At Sumus, I separated what the MVP needed to prove from future ambitions and reworked the plan around the startup's runway. Planned development cost and time to market fell by 85% each. I presented the roadmap, budget and trade-offs to the founder before committing to agency development.
At Skillward, my responsibilities extended from building the platform into product planning, customer acquisition, sales, marketing and fundraising.
Read the Sumus storyEngineering practice
Architecture and delivery processes need attention together.
Hiring, development standards, deployment and quality checks all form part of how software gets built. My work has included establishing those practices as well as designing the software.
At Blackford, I established a delivery function for an art marketplace, recruiting seven people in four weeks. I designed the architecture and delivery setup around developer onboarding, security review, design testing and visibility for the client-facing team.
At Appraisd, I introduced Scrum across engineering and product, a process for production bugs and client requests, and static analysis and automation in the pull request and CI/CD pipeline.
Read the Blackford storyAI-assisted engineering
AI is part of how I build and review software.
At Blackford, AI-assisted coding and pull request reviews were part of a delivery setup built around microservices, automated test harnesses and CI/CD.
At Responsible Tech, I use spec-driven development (SDD), alongside iterative implementation, testing, AI evaluations, CI/CD and documentation.
For Vertie, I separated the system by discipline and kept backend modules decoupled so the work remained manageable within an AI coder's context. Tests and evaluation harnesses were established from the start.
Applied AI
Model selection is part of the engineering work.
I want model choices to be grounded in the task the product needs to perform. Retrieval, evaluation and operating cost belong in that discussion alongside the user experience.
At Skillward, I built a coach agent that used coach-uploaded resources through RAG and tools to manage goals and progress. I evaluated language models using synthetic benchmarks for task performance, cost and response accuracy.
Redraft explores a different part of the AI experience: a background reader that surfaces story facts and continuity issues while the writer works.
Explore Skillward's AI platformData and user control
Where the work lives is a product decision.
I'm interested in software that gives people more control over their work. In my recent projects, that includes browser storage, files in a user's own GitHub repository and AI features that use their own provider key.
Manatee stores diagrams as Mermaid files and offers optional GitHub sync, with a diff to review AI changes before applying them. Redraft offers local drafts, optional GitHub sync and exports to Markdown, plain text and PDF.
See the projectsLet's talk about your technical direction and engineering plans.
Get in touch