Senior ML Engineer - LLMOps & Trustworthy GenAI
Production GenAI · LLM evaluation · RAG systems · AIOps · ML security & privacy
I build and operate ML systems for reliability- and safety-critical environments: evaluation pipelines, RAG services, model-serving infrastructure, observability, safety/privacy guardrails, and AI-based reliability tooling for large-scale systems.
This GitHub profile page is a curated summary of my public work, including selected demos, reproducible experiments, OSS contributions, and links to case studies.
- LLMOps & GenAI platforms: model evaluation, CI quality gates for GenAI models, RAG, agents, LLM serving and inference
- Production ML infrastructure: observability, SLOs, Grafana/ELK, CI/CD, model and dataset versioning
- AIOps & reliability: online failure prediction, root-cause analysis, anomaly detection, operational risk analysis and prevention
- ML security & privacy: guardrails, confidential computing, TEEs, differential privacy, model-risk evaluation
- Applied ML: NLP, computer vision, graph learning, reinforcement learning, tabular models
| Area | Public artifact | Why it matters |
|---|---|---|
| Portfolio & case studies | paolo-notaro.github.io | Long-form overview of projects, publications, and technical writing |
| LLM / data app demo | NaturalQL | Natural-language-to-SQL demo with schema-bounded prompting and safe execution |
| Computer vision | chess-detector | Vision-only chess move tracking from raw camera images |
| Research | Publications | AIOps, failure management, RCA, command-risk classification, reliability |
| Blog | Technical notes | Write-ups on ML demos, RL experiments, and applied AI projects |
- Building production-style public demos around evaluation, RAG, and LLM application safety.
- NaturalQL: Natural-language-to-SQL demo with schema-bounded prompting and safe execution
- Professional work includes LLM evaluation platforms, CI quality gates, RAG services, vLLM/Kubernetes serving, and GenAI observability. Public case studies and simplified demos are linked from my portfolio.
- PhD research on AI-based proactive failure management for large-scale cloud systems. See here.
- First-author publications on AIOps surveys, structured-log root-cause analysis, and operational risk classification.
- Work on command-risk classification, guardrails, confidential computing, and privacy-preserving ML.
- Lead inventor on a command-auditing patent for operational-security workflows.
Selected external contributions:
code-yeongyu/oh-my-openagent— PR #3802: prevented an infinite loop when the fallback model equals the current model.opencode-code-quota— PR: contribution to quota / coding-agent tooling.
I also use personal repositories to document engineering practice: packaging, CI, documentation, tests, reproducible demos, and public technical write-ups.
- Languages: Python, C++, SQL, Bash, Java
- LLM / GenAI: vLLM, Hugging Face Transformers, LangChain/LangGraph, agents, MCP, FAISS, ChromaDB, hybrid retrieval
- LLMOps / MLOps: MLflow, DVC, W&B, MinIO, Docker, Kubernetes, Helm, ArgoCD, FastAPI
- Observability: Prometheus, Grafana, ELK
- ML / DL: PyTorch, PyG, scikit-learn, OpenCV, Gym
- Quality: GitHub Actions, GitLab CI/CD, Poetry, uv, Ruff, pytest, mkdocs
- Security & governance: EU AI Act, GDPR, model-risk evaluation, guardrails, TEEs, differential privacy
- Portfolio: paolo-notaro.github.io
- Publications: paolo-notaro.github.io/publications
- Blog: paolo-notaro.github.io/blog
- LinkedIn: linkedin.com/in/paolo-notaro
I am especially interested in production-grade GenAI, LLM evaluation, AI reliability, model-risk management, and ML systems that need to work under real operational constraints.