I add AI to software that already exists, and automate the manual work around it. Not demos — features that run in production, handle bad input and cost you a predictable amount per month.
What I build:
- AI features inside your app or website: summarisation, classification, extraction, drafting, search and Q&A over your own documents
- Chatbots and assistants trained on your content, wired into your database or knowledge base
- Document and data extraction: invoices, PDFs, forms and scanned files turned into structured data
- Workflow automation: reports, data syncs between systems, scheduled jobs and repetitive back-office tasks removed
- Machine learning models in Python for classification and prediction, with honest evaluation of whether the accuracy is good enough to ship
- API-first design so the AI piece stays swappable and you are not locked to one vendor
Tech: OpenAI and Claude APIs, prompt design, retrieval over your own documents, Python, scikit-learn, pandas, NumPy, C# / .NET 8, ASP.NET Core Web API, Azure Functions, MS SQL Server, REST, JSON, webhooks.
Background: alongside my .NET development work I have done applied machine learning research, including a signal-classification benchmark comparing multiple model architectures — so I can tell you when a problem genuinely needs a model and when a few rules and a good API call will do the job for a fraction of the cost.
How I work: we start with the specific task you want handled, I build the smallest version that proves it works, then we expand. You get clear notes on running costs before anything is built, so there are no surprise bills.
Have a repetitive process eating your team's time? Describe it and I will tell you whether it is worth automating and roughly what it would take.