Description

Senior AI/ ML Engineer and Technical Leader with 9+ years shipping production ML/ LLM systems at massive scale. Deep hands-on expertise in PyTorch, the modern LLM stack (vLLM, HuggingFace, fine-tuning pipelines, RLHF/ DPO, eval frameworks), and inference optimization — consistently delivering sub-100ms latency for 100K+ daily users. Built and led teams of 10+ engineers, co-founded an AI startup as CTO, and shipped 12+ AI features from concept to production in rapid sprint cycles. Equally comfortable architecting systems, writing production Python code, and mentoring engineers. Thrive in high-autonomy, high-velocity environments where shipping fast and iterating on real user impact matters more than perfect academic solutions.

Languages

English
Native or bilingual
Italian
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

Nike, Nike
Senior AI/ML Engineer - Senior Full Stack Engineer
June 2025 - January 2026 (7 months)
Via Milan, San Giorgio Delle Pertiche, PD, Italy
• •Architected LLM-based agentic workflows using LangGraph for multi-step autonomous reasoning, enabling self-service troubleshooting across 5 product engineering teams serving 50K+ daily active users
• • Designed and deployed advanced RAG pipelines with hybrid retrieval (dense + sparse), reranking, and query decomposition, improving search relevance by 34% over previous implementation
• • Led inference optimization initiative using vLLM and TensorRT, reducing LLM serving costs by 45% while maintaining <100ms p99 latency at scale
• • Implemented LLM evaluation and monitoring framework with Weights & Biases for automated accuracy tracking, hallucination detection, and model performance drift across production workloads
• • Built real-time WebSocket streaming dashboard for LLM performance monitoring, prompt analytics, and usage metrics across engineering teams
Machine learning Node.js artificial intelligence LLM Python
Loro Piana (LVMH), Loro Piana (LVMH)
Tech Lead / Senior AI Engineer
January 2025 - July 2025 (6 months)
Milan, Metropolitan City of Milan, Italy
• • Led team of 10 engineers (backend, frontend, ML) building AI-powered platform processing €2M+ daily transactions across 50+ countries
• •Architected NLP-based entity resolution pipeline using PyTorch and Hugging Face Transformers for intelligent matching and deduplication across multi language datasets, reducing duplicates by 73%
• • Designed PySpark pipelines on GCP Dataproc for large-scale data processing and feature engineering across distributed global datasets (5TB+ monthly throughput)
• • Established MLOps framework with MLflow model registry, automated A/B testing, and Datadog monitoring for production model performance tracking
• • Delivered end-to-end MVP to production in 4 weeks using rapid iteration methodology with CI/CD pipelines — shipping fast in ambiguous environments
& CTO, Lemnia
Co-Founder
August 2024 - Today (1 year and 10 months)
Milan, Metropolitan City of Milan, Italy
• •Architected end-to-end NLP pipeline using custom PyTorch models converting natural language to executable API workflows, achieving 94% intent classification accuracy — demonstrating startup-speed execution and owner mindset
• • Built production RAG system with LLM orchestration (LangChain, LangGraph) and vector databases (FAISS, pgvector) for semantic retrieval, reducing manual configuration by 80%
• • Deployed custom PyTorch models for sequence labeling, coreference resolution, and entity extraction with ONNX Runtime inference optimization achieving sub-200ms latency
• • Implemented reinforcement learning pipeline in PyTorch for self-optimizing workflow generation, with MLflow experiment tracking and Weights & Biases for training monitoring
• • Designed agentic AI system with multi-step reasoning and autonomous tool use, parsing Swagger/OpenAPI specs and introspecting database schemas for zero-config automation