UK finance professional · AI engineer
A finance professional who ships real AI: accounting automation, LLM evaluation, and agent tooling. Most AI builders cannot do finance, and most finance people cannot build. I do both - and the work is adopted, shipped, and measured.
Who you would be working with
I am a UK finance professional running automation and AI at a UK accountancy practice, and I build and ship production AI systems. I hold a BSc in Banking and International Finance from Bayes Business School.
What makes the work different is the overlap: I bridge deep UK financial domain knowledge - accounting, VAT, HMRC, double-entry, payroll - with hands-on AI engineering. That means I can take a messy finance problem, model it correctly, and build the system that solves it. I am a published open-source author and a hackathon-winning builder.
What I can build for you
Four things I deliver end to end, where deep UK financial domain knowledge and hands-on AI engineering are both required.
Transaction categorisation, compliance tracking, and Companies House API integration - grounded in real double-entry, VAT, and the UK Chart of Accounts that sits behind Sage, Xero, and HMRC filings.
End-to-end systems and integrations across OpenAI, Anthropic, Gemini, Bedrock, and Ollama - including agent tooling, OCR pipelines, and structured, auditable outputs.
Benchmarking, evaluation harnesses, and adversarial testing - with a rare specialism in the finance and accounting domain, where correctness is not optional.
React and TypeScript on the front, Python and FastAPI on the back, AWS serverless underneath. Shipped to the web, to iOS, and to live production APIs.
Each piece answers one question: what could I deliver for you, and what proves I can?
Open source · adopted in production
A zero-dependency Python library providing the complete UK Chart of Accounts: 166 nominal codes, VAT treatments, HMRC box mappings, double-entry rules, and an LLM prompt export. It fills a genuine gap in the Python ecosystem - and real developers use it.
End-to-end AI system, inside finance
A multi-client AI bookkeeping system: LLM transaction categorisation against the UK Chart of Accounts, OCR receipt processing, confidence scoring with reasoning for audit trails, a double-entry ledger, and isolated client workspaces. Built on AWS with Bedrock, DynamoDB, S3, and Cognito.
LLM evaluation · specialist domain
I authored and submitted a UK bookkeeping benchmark to Google DeepMind's "Measuring Progress Toward AGI" competition (Kaggle Community Benchmarks, Learning track). It probes where frontier models hold up - and where they quietly break - on real UK accounting tasks.
Automation in a live practice
As Head of Systems & Automation at a UK accountancy practice, I built an automated compliance-tracking dashboard (Python plus the Companies House API) covering confirmation statements, statutory accounts, and corporation-tax deadlines; an ML-powered bookkeeping categoriser against UK nominal codes (rule matching, scikit-learn, and LLM inference); and a multi-provider desktop AI agent.
Shipped end to end, at scale
A property environmental-quality tool and B2B API: noise, air quality, and liveability scoring across London neighbourhoods, combining 10+ live government data sources. Fully serverless on AWS, with a live API and an iOS app shipped to the App Store.
Also shipped: Noor - a multi-platform consumer app (15 modules, right-to-left and multi-language support, offline-first, full CI/CD across web, iOS, and Android); and Sterling - a voice-first assistant entered into the Gemini 3 Hackathon (~34,000 participants).
See it, don't just read it
This is the judgement my tools automate. Pick the UK nominal code - the same taxonomy my uk-chart-of-accounts library encodes for machines to use.
Monthly office rent paid via BACS
£1,200
Which nominal code?
reasoning Property rental is a recurring operating overhead - expensed to the P&L as incurred, never capitalised.
Available now
I take on contract and freelance work in finance automation, AI and LLM development, and LLM evaluation. If that is what you need, get in touch. I read every serious enquiry and reply within a day - no pitch, just an honest answer on whether I can help.
Or reach me directly: bilalkhizar@hotmail.co.uk