I am an AI Engineer fascinated by the latest research. I spend my time learning, experimenting, and bringing new ideas to life.
I currently lead the technical delivery of AI solutions at BIP / Datwave, transforming theoretical breakthroughs into enterprise-grade features.
- Agentic Frameworks: Architecting low-latency (~800ms) Reasoning-Action loops and multi-agent orchestration using Google’s ADK.
- Reliability & Alignment: Implementing self-consistency loops and rigorous guardrails that reduced system hallucination rates from ~5% to near-zero.
- Production MLOps: Deploying end-to-end vision and forecasting pipelines on GCP in high-stakes domains.
Also working on improving my GitHub profile! Stay tuned.
- NeurIPS 2024: First author of "Sub-optimal Experts mitigate Ambiguity in Inverse Reinforcement Learning". My research demonstrates how multiple sub-optimal demonstrations can significantly shrink the feasible reward set in IRL.
- Current Interests: Mechanistic Interpretability, LLM post-training, and AI Alignment.
- Deep Learning: PyTorch, vLLM, Transformers, CNNs.
- Engineering: Python, FastAPI, Pydantic, ADK, React, UV, Ruff.
- Cloud & MLOps: Google Cloud Certified Professional ML Engineer (Vertex AI, BigQuery, Cloud Build, Terraform, Docker).
- Outdoors: Avid hiker and photographer. Always happy to share recommendations for the best trails and viewpoints in Northern Italy.
- Gaming: Videogame enthusiast, always happy to talk about latest releases.
Invited Speaker at Google GDG Zürich | Active Hackathon Participant