SELECTED WORK
A selection of projects where off-the-shelf tools weren't good enough and something custom had to be built.
Professional Services — Confidential
A custom client management platform built entirely on the client's own servers. No cloud. No third-party access. Every record stays in-house.
The client needed a single system to manage sensitive client relationships across multiple teams. Every SaaS tool they looked at was a non-starter — the data couldn't leave the building. They also needed it fast: any query, on any record, had to return in under 10 milliseconds.
We built the whole thing from scratch and installed it on their own hardware. The database was tuned specifically for how they search and filter their data. Access controls were built to be granular — different staff see different fields depending on their role. The interface was designed around their actual daily workflows, not a generic template.
Retail & Distribution — Confidential
A full e-commerce platform built from the ground up — product catalogue, pricing, orders, fulfilment, and post-sale operations — all in one owned codebase.
The client had outgrown every platform they tried. Their pricing rules were too complex, their fulfilment logic too specific, and their inventory system too custom for any off-the-shelf solution to handle. They were paying for three separate tools that barely talked to each other and breaking under peak load.
We built each part of the commerce process as its own service — catalogue, pricing, orders, fulfilment — so they could be updated independently. The pricing engine evaluates complex rule combinations in real time. The order system was built so that if anything fails mid-process, it recovers cleanly without creating duplicate or broken orders.
Legal — Confidential
An AI assistant built to run entirely inside the firm's own network. It searches documents, flags deadline risks, and drafts internal summaries — without sending a single word to an outside server.
Senior associates were spending hours on tasks that didn't need their expertise — cross-referencing documents, checking matter status, writing internal briefings. The firm wanted AI to handle that work, but client confidentiality meant nothing could go near an external API.
We deployed a language model on the firm's own servers and connected it to their document and matter management systems. Staff can ask it questions in plain English and get answers with source references. Every response is logged so there's a clear record of what the AI said and why.