Selected work in pharmaceuticals, energy and financeGeorgios Vamvouras · 20+ peer-reviewed publications · 5+ clients
AthensPausedContents
WinMedica
Pharmaceutical
Every word on the pack checked against the approved text.
Leaflets, cartons and blisters, aligned word by word, every difference marked where it is. In daily use by three departments.
The LabArt pageVP&C
Pharmaceutical policy
A digest built around the reader’s interests, every item checked against its source.
In use by large pharmaceutical companies and national policy-makers.
WinMedica
Energy
Sized and costed end to end, using methods from our published work.
Weather, price and load forecasting, energy yield, battery dispatch, feasibility and lifecycle analysis. Efficiency, reliability and revenue, each quantified.
Lab.RACSE, NTUA
Energy
82% of the theoretical maximum.
The maximum is what a battery would earn if it knew tomorrow’s prices and had the perfect plan. Our learned controller comes that close with real forecasts, and outperforms all the popular industry methodologies.
Lab.RACSE, NTUA
Energy
Scored on a full year the models never saw.
Deep-learning forecasters for hourly prices and 48-hour temperature, irradiance and humidity, published. The same forecasts powered the analysis of WinMedica’s Tripoli plant (chapter 3).
A private investor
Finance
Most of the stock market’s growth, with about half the fall in a crash.
Eight index funds across world stocks, bonds and gold, rebalanced automatically every quarter: the calmer funds get more of the money, and no fund gets more than 40%. In historical tests it has earned about 9.6% a year since 2012, close to a stock-heavy portfolio, and in 2008 it lost about half as much. It makes no predictions and never times the market; before each rebalance the investor gets a notice and can decline.
Stathmon
In-house product
An assistant that gets to know you.
Tell it what you need, by voice or in writing. It keeps your tasks and reminders, reads what you share with it, learns how you work and answers from what it knows about you. Access it from your phone or the web.
The Right Hand pageLocal detectors trained on the client’s own documents find every word on pharmaceutical artwork, more precisely than the strongest commercial alternative.
WinMedicaReads Greek and English pack text more accurately than the commercial cloud OCR services. Built for regulatory documents.
WinMedicaStructured fields from scanned invoices of varied layouts. Validated and commercialised.
VisionITOpen-weight models deployed and benchmarked on your own GPUs or HPC allocation, so data stays on-premise. Depending on the task, they reach 80–100% of the capability of the best closed frontier models.
Regression models that predict cycle performance, including fluids not seen in training.
Lab.RACSE, NTUAPlugins for building energy software such as IDA ICE that run automated design optimisation on the model.
Lab.RACSE, NTUAWe start from the workflow, the data and the decision, with the people who will use the system, and agree what will be measured.
We build on your own data, with published methods or our own research, and test against the cases that matter to you before anything is used.
The system runs on your infrastructure or ours, with documentation, training and continuing support.
Founder and principal engineer. PhD candidate in Energy and AI; MSc Data Science and Machine Learning (2027); Dipl.-Ing. Mechanical Engineering. National Technical University of Athens.
Personal site and publicationsTell us the workflow, data or decision you want to create or improve.