Selected independent work
Projects built to be explored.
Public tools, technical experiments and open-source contributions. The visuals below come from the working demos themselves—open one to try the product, or inspect the source where it is available.
- 8
- public projects
- 6
- live products & demos
- 1
- open-source contribution

Residue Lens
Built and launched a Next.js/TypeScript product with Python import and validation pipelines that keep official UK, US and European pesticide-monitoring programmes comparable without pooling unlike evidence.

Period Place
Built and launched a Next.js/TypeScript UK finder backed by libSQL/Turso, with MapLibre search, moderated community reports, audit history, guarded environment operations and explicit evidence gaps.

Budget V2
Built a Python/PostgreSQL personal-finance ledger with immutable source evidence, idempotent provider refreshes, deterministic classification and reconciliation, guarded monthly delivery and verified backup/restore; demonstrated publicly with synthetic data.

Backtest Library
Extended a Python backtesting library with filing-date-aware fundamental signals, a matched comparator and a predeclared data-quality audit using derived StockFit trial results.

Dependency Agent
Built a bounded Python/FastAPI developer tool that reads Java and Maven source evidence, traces a compilation failure to an incompatible SLF4J version and produces the smallest verified POM patch.

Strava Dashboard
Built a demo-first training dashboard with Strava OAuth, interactive activity analysis and OpenAI-generated coaching insights, while preserving a private live-data mode.

Pulse Energy Dashboard
Built a privacy-first dashboard that discovers Octopus smart meters, turns half-hourly electricity and gas readings into clear usage patterns, estimates tariff costs and offers optional server-side AI coaching.

Computational Geometry CAD Filter
Extended a shrinking-ball medial-axis pipeline to expose geometric scale features from sampled CAD surfaces, then rebuilt it with reproducible C++ tooling and an interactive browser reconstruction.
Ongoing coding practice
LeetCode practice
Regular algorithm and data-structure practice across Java, Python and Pandas, building fluency through consistent problem solving.
- Easy
- 83
- Medium
- 57
- Hard
- 3
Languages
- Java 83
- Pandas 43
- Python3 17
Practised areas
- Array 56
- Hash Table 21
- Two Pointers 20
- Depth-First Search 15
- Dynamic Programming 12