AFAndrei Foitoș
01

Insights Dashboard — Neolook

Full-stack web application to monitor and analyse data from neonatal intensive care units, built in a collaborative, company-like environment.

Deployed and actively used in production by Neolook in real clinical settings.

ReactDjangoRESTJWT
02

Epic Serv CRM

Internal CRM for a gas-installation company — a document-per-job workflow with file management, role-based access, and cloud document storage. Designed, built, and deployed end to end.

Shipped to production and in daily use. FastAPI + Next.js, Supabase, Google Drive integration, JWT authentication.

Next.jsFastAPIPostgreSQLGoogle Drive API
039.5 / 10

O₃ & NO₂ Air Quality Forecasting

Forecasting model predicting ozone and nitrogen dioxide concentrations three days ahead, with a user-facing interface based on WHO guidelines.

Residents can check localised air quality forecasts for informed daily decisions.

PythonMachine LearningTime Series
049 / 10

Disentangling Uncertainty in Facial Age Estimation

“Disentangling Model and Human Data Uncertainty in Apparent Facial Age Estimation” — with Ivo de Jong and Matias Valdenegro-Toro (University of Groningen). Trains Bayesian neural networks (MC-DropConnect, Flipout, Deep Ensembles) to separate model uncertainty from human-label noise, and probes it under distribution shift.

Accepted for oral presentation at ANNPR 2026 (IAPR TC3 Workshop), to appear in the Springer LNAI proceedings. Bachelor's thesis graded 9/10.

PyTorchDenseNet121Bayesian Neural Networks
058.5 / 10

DQN with Temporal Decay + PER

Extended a Dueling DQN with Prioritised Experience Replay by introducing an exponential temporal decay factor over transition age.

Demonstrated improved sample efficiency — recent transitions are weighted more, without discarding rare but informative events.

Reinforcement LearningPyTorchRL Theory
068 / 10

Drum Pattern Generation via FNN

Designed and trained a Feedforward Neural Network to generate drum sequences conditioned on MIDI seed data.

Produced rhythmically coherent and musically plausible patterns; explored music as a structured sequence prediction problem.

PythonNeural NetworksMIDIMusic AI