Log

Just finished a four-day KTP residential training at Ashorne Hill, near Leamington Spa. It brings KTP associates from across the UK together and focuses on project management, leadership, presenting and team development. A brilliant opportunity to meet other associates and learn more about myself: how I work in a team, how I collaborate, and the areas I want to improve.

LiftPad: shipped my personal gym tracker for an Upper/Lower/Push/Pull/Legs routine. Works fully offline via localStorage with optional Neon Postgres sync, and exercise demos open straight to YouTube. Live on Vercel.

KTP: ran a multimodal pain-classification study on the BioVid dataset — comparing the existing facial-landmark video features against physiological signals (EDA, ECG, EMG) using SVM and Random Forest, with leave-one-subject-out cross-validation on the university HPC cluster.

Almanac: built a personal command-centre dashboard — a 'morning briefing' agent that pulls my calendar, email, bank statements and Trading 212 portfolio into one Next.js + Postgres view, with daily AI-generated summaries. Cron-driven ingest keeps it fresh. Runs locally, just for me.

Built a quick landing page for a friend's property management business in Germany (Klein & Fein Hausmeisterservice). Fun side project, nice to help people get online with a clean, professional site.

Foundry 70: updated landing page and website based on feedback from early conversations. Expanding the network, connecting with local business communities and exploring partnerships in Hertfordshire.

Foundry 70 is picking up pace. Building this business alongside the KTP role. Two jobs, one goal: ship things that matter.

Started a new role: Research Scientist on a Knowledge Transfer Partnership (KTP) between the University of Hertfordshire and a pharmaceutical research company. KTPs are government-funded partnerships where universities and businesses collaborate on innovation projects. I manage the full project end-to-end, bridging both the university and the company, driving the machine learning research and applying AI to real-world pharmaceutical data.

Jigsaw Puzzle Generator: full training infrastructure done. Siamese Neural Network architecture presets, auto PyTorch script generation, local + cloud (SSH) training with live progress, experiment tracking with script hashes. Ready to start training runs.

Updated project page with all 6 phases. Phases 1-3 complete (data gen, dataset gen, training infra). Next: train the actual piece matching model.

Jigsaw Puzzle Generator: added Openverse image fetching (tens of thousands of images), auto-normalization for ML training, and puzzle cutting at any dimension. Posted a video demo on X. Open source on GitHub.

Added my personal story to the site. About page is live.

Built a jigsaw puzzle generator (Swift + SwiftUI). Open sourced it. First step toward the puzzle-solving robot project.

OpenAnalytics: public demo dashboard live, city-level geo map, WS auth ticket flow, landing page redesign.

Started planning the world's fastest jigsaw puzzle solving AI robot. Researched state of the art, defined 6 possible paths.

LinkedIn Dashboard: added X post history endpoint, fixed draft ordering, integrated analytics tracking.

Launched ClawPost on RankInPublic. Shared my story replying to Simon Squibb's post (7 likes, 432 impressions).

Started 30-day X posting challenge. Day 1: committed to showing up every day.