Skip to content

cat work.log

Built, tested and live

Each project shows the problem, what we built and the measured results. Every demo is open to the public and the code is on GitHub.

Projects

Kendot Legal

liveconcept buildLegal

A law firm website with a cited AI assistant for visitors, a private one for lawyers, and paid consultation booking from Google Calendar.

100%
Answer correctness on a 40-question test set
65/65
Citations checked against their source page
100%
Prompt-injection attempts resisted
  • Astro
  • Tailwind
  • FastAPI
  • Claude Haiku 4.5
  • Voyage embeddings
  • Postgres + pgvector

case study →: Kendot Legallive demo ↗code ↗

FlyFit Fashion Store

liveconcept buildRetail and e-commerce

An online fashion store with checkout that can't oversell, payments that survive duplicate webhooks, AI search and a shopping assistant that can't make up a price.

0
Items oversold with 200 buyers racing for 10 units
98.3%
Shopping assistant tasks completed (40 scenarios, 3 runs)
0
Made-up prices shown, audited against the database
  • Python
  • FastAPI
  • Postgres + pgvector
  • Next.js
  • TypeScript
  • Tailwind

case study →: FlyFit Fashion Storelive demo ↗code ↗

Naija Law RAG

liveLegal information

An assistant that answers questions on Nigerian law strictly from the official texts, cites every claim, and declines what the documents don't cover.

11/11
Citations verified against the source text
5/5
Out-of-scope questions correctly declined
8/9
Answerable questions answered correctly
  • Python
  • FastAPI
  • Claude
  • Voyage embeddings
  • BM25
  • Qdrant

case study →: Naija Law RAGlive demo ↗code ↗

Helpdesk RAG

liveIT support

An IT helpdesk assistant that answers staff questions from the company's own knowledge base, with a source for every claim and a clean "I don't know" for everything else.

6/6
Answerable questions answered correctly
18/18
Citations verified against the source
5/5
Out-of-scope questions declined
  • Python
  • FastAPI
  • Claude
  • Voyage embeddings
  • BM25
  • Postgres + pgvector

case study →: Helpdesk RAGlive demo ↗code ↗

Football Match Predictor

liveSports analytics

A machine learning model that gives win, draw and loss probabilities for matches in seven European leagues, tested honestly on matches it never saw, and retrained every week.

50.7%
Results called correctly on 1,043 unseen matches
+8 pts
Better than always picking the home team (42.8%)
0.999
Log loss, against 1.076 for the baseline (lower is better)
  • Python
  • pandas
  • scikit-learn
  • PyTorch
  • Weights & Biases
  • MySQL

case study →: Football Match Predictorlive demo ↗code ↗

cat next-step.md

Want something like this for your business?

Book a free 30-minute strategy call. You'll leave with at least one concrete idea, whether or not we work together.

Book a Free Strategy Call