ML Engineer · Vienna, Austria
I build ML systems end-to-end — from data pipelines and model training to containerised deployment. Four years at IST Austria and IIT Madras. Ask me anything.
Selected work
Deep autoencoder learning compact spatial representations from 250k 3D trajectory samples. Task-relevant structure emerges without explicit supervision — validated across two experimental paradigms.
Multi-task CNN–GNN trained on 1M+ synthetic trajectory points via a custom Unity3D pipeline. 7-config ablation on prediction targets; EWC continual learning across 7 sequential sessions.
End-to-end RL pipeline — custom 49-neuron state encoder at >90% position reconstruction, TD-trained value network, Go-Explore-NoGo paradigm for hard exploration. No explicit path planning.
Playwright scraper + Flask backend + Claude API scoring dashboard. Dark-themed UI with salary tabs, Applied Jobs tracking, server-side persistence, CSV export, and profession-agnostic scoring prompts.