Founder & Principal Engineer Codgent
- Consultancy practice: takes products from architecture to deployment for startups and enterprises as a hands-on engineer (after briefly scaling the agency) on Python, Django, TypeScript, Node.js, Next.js and AWS, with Docker isolation, automated zero-downtime releases and responsible agentic development.
- Wild Me (Conservation X Labs) - serverless AWS migration: helped move the legacy ML wildlife identification pipeline, which powers automated animal identification across 53+ species, to serverless AWS (Chalice, on-demand ML compute, Lambda, pgvector), cutting cost about 25x and runtime 150x; worked with the ML team on the integration points between the application and their computer vision matching, while maintaining the legacy Java platform it belongs to, which serves tens of thousands of users; built and open-sourced WildbookExport so researchers can export data for offline analysis.
- AutoValuator - data ingestion and valuation engine (Python, Scrapy, PostgreSQL/PostGIS, AWS, Docker, Next.js): designed the production pipeline that sources live vehicle listings, normalises and deduplicates them by VIN across sources, and applies radius, mileage and trim match rules; replacing scraping with API sourcing cut a comparable set from roughly 20 minutes to a handful of seconds and reduced sourcing failures by about 80%.
- AutoValuator - report parsing: built the parser that turns CCC One insurance valuation reports (PDF and Excel) into structured case data (VIN, mileage, options, insurer adjustments and the insurer comparables), with OCR fallback for scanned pages and logic that survives layout drift between report revisions; total-loss and diminished-value reports render through a custom templating engine, and the client received a video walkthrough and written documentation on updating parsing rules and data sources.