Yekaterina Donegal
Yekaterina Donegal
Hey there 👋

I'm Yekaterina Donegal

Data Scientist · Pace University MS · NYC

Data science grad student at Pace with backgrounds in psychology and anthropology. I work on applied ML for understanding behavior and decision-making — multimodal models, time-series forecasting, and retrieval-augmented systems. This site is where I keep my projects and writing.

Get in touch GitHub

Selected projects

All projects →
CAMELS — Multimodal latent space for human–AI interaction

CAMELS — Multimodal latent space for human–AI interaction
Team · Research

Research project building a real-time multimodal understanding pipeline that fuses what someone is doing on video, the phonemes they're producing, and the prosody of their voice into a single shared latent space, so a conversational AI can react to all three jointly.
Life Expectancy — country-grouped cross-validation

Life Expectancy — country-grouped cross-validation
Submitted

When does a 0.95 R² mean your model works, and when does it mean your validation leaked? A WDI panel-data study that started as a coursework project and turned into a methodological correction submitted to IEEE GHTC 2026.
Restaurant Allergen Assistant

Restaurant Allergen Assistant

Real-time allergen lookup for restaurant staff. Combines OCR for menu ingestion, GPT-3.5 for reasoning, and Retrieval Augmented Generation for grounded allergen lookups.

Latest writing

All posts →