Skip to content
← Research
Applied Research

Scientific AI

Machine learning built for scientific instruments — where correctness, calibration, and physical constraints matter as much as accuracy.

AutoencodersLinformerLatent-space diagnosticsAnomaly detection

A research thread carried out during a research internship with CERN openlab's Next Generation Triggers initiative: building PyTorch representation-learning models for LHC Level-1 trigger event embeddings, evaluating Linformer and multi-head attention architectures for HH4B-vs-Standard-Model anomaly detection, and diagnosing latent-space behavior with PCA, effective rank, Mahalanobis scores, and Fisher analysis.