[ Projects ] · 9 selected case studies
Projects
A selection of projects: what was built, how it was evaluated and what it changed. Client names stay private; every case study shows its architecture.
- 01LLM agent that turns IT support requests into ticketsAn agent that reads support emails and call summaries and opens, categorises and updates tickets in the support team's ticketing tool, with code checks around every model step.
- 02Platform for reviewing and flagging online contentA platform that helps operators review content found online, report it to the platforms that host it and follow up on each report.
- 03Coding agent for loading files whose layout keeps changingPartner files arrive by email and their layout drifts. A known layout runs an approved loading script with no model call; a new one goes to a sandboxed coding agent, then to automated checks and a human approval.
- 04Agent-assisted migration of legacy ETL jobs to PySparkLegacy ETL jobs and master-data (MDM) flows rebuilt one for one in PySpark. Plain code writes the skeleton, an agent fills only the marked gaps, and a job is accepted when its outputs match today's.
- 05Assistant that answers from internal documents (RAG)A chat assistant that answers from internal document libraries and optional web search, cites its sources and can return a Word report.
- 06Invoice classification with Blue/Green releasesRetraining, support and Blue/Green releases (two identical deployments, switched in one step) for an invoice classifier used daily by a finance team.
- 07Predictive purchase models in the data warehouseSeven models of what drives customers to buy, built and scored inside the business's data warehouse, with automated deployments and indicators for analysts.
- 08CV extraction and search by meaningAn internal tool that extracts structured profiles from CVs with OCR (text recognition) and an LLM, then finds matches by meaning with an explanation for each.
- 09Tagging algorithm problems from their statement and codeA classifier that assigns one or more of eight algorithm tags to a programming problem, from its statement and a Python solution, compared with an LLM given no examples.