Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m Diego Mazzalovo, a Data & AI Engineer with 4+ years of experience across the full machine learning lifecycle—from data engineering and model training to deployment and production monitoring. I enjoy turning messy real-world data into reliable systems, using PyTorch and modern MLOps practices with Azure and Docker. More recently, I’ve been focused on LLM-powered solutions, including local SLM fine-tuning, GenAI-based classification workflows, and RAG prototyping with vector databases. I like building systems that are both accurate and practical—using structured outputs, validation layers, and iterative improvements based on performance signals from production.…I’m Diego Mazzalovo, a Data & AI Engineer with 4+ years of experience across the full machine learning lifecycle—from data engineering and model training to deployment and production monitoring. I enjoy turning messy real-world data into reliable systems, using PyTorch and modern MLOps practices with Azure and Docker. More recently, I’ve been focused on LLM-powered solutions, including local SLM fine-tuning, GenAI-based classification workflows, and RAG prototyping with vector databases. I like building systems that are both accurate and practical—using structured outputs, validation layers, and iterative improvements based on performance signals from production.

Diego Mazzalovo

AI Engineer, AI Developer, Data Scientist





I’m Diego Mazzalovo, a Data & AI Engineer with 4+ years of experience across the full machine learning lifecycle—from data engineering and model training to deployment and production monitoring. I enjoy turning messy real-world data into reliable systems, using PyTorch and modern MLOps practices with Azure and Docker. More recently, I’ve been focused on LLM-powered solutions, including local SLM fine-tuning, GenAI-based classification workflows, and RAG prototyping with vector databases. I like building systems that are both accurate and practical—using structured outputs, validation layers, and iterative improvements based on performance signals from production.…I’m Diego Mazzalovo, a Data & AI Engineer with 4+ years of experience across the full machine learning lifecycle—from data engineering and model training to deployment and production monitoring. I enjoy turning messy real-world data into reliable systems, using PyTorch and modern MLOps practices with Azure and Docker. More recently, I’ve been focused on LLM-powered solutions, including local SLM fine-tuning, GenAI-based classification workflows, and RAG prototyping with vector databases. I like building systems that are both accurate and practical—using structured outputs, validation layers, and iterative improvements based on performance signals from production.

Available to hire

I’m Diego Mazzalovo, a Data & AI Engineer with 4+ years of experience across the full machine learning lifecycle—from data engineering and model training to deployment and production monitoring. I enjoy turning messy real-world data into reliable systems, using PyTorch and modern MLOps practices with Azure and Docker.

More recently, I’ve been focused on LLM-powered solutions, including local SLM fine-tuning, GenAI-based classification workflows, and RAG prototyping with vector databases. I like building systems that are both accurate and practical—using structured outputs, validation layers, and iterative improvements based on performance signals from production.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Beginner

Language

Italian

Fluent

English

Advanced

Work Experience

Data & AI Engineer at Capgemini Italia

September 1, 2024 - Present

At Capgemini Italia, I build and improve end-to-end ML/AI solutions spanning LLM systems, RAG, and computer vision. I developed a domain-specific RAG client prototype by fine-tuning a local Gemma3 SLM with LoRA (via Unsloth/SEFT), achieving a ~10% performance lift on complex, field-specific technical Q&A. I also implemented an LLM-driven resume screening pipeline using multi-step reasoning, structured outputs with Pydantic, and validation layers, reaching ~80% accuracy for technical role matching while keeping the evaluation explainable. On the computer vision side, I improved production performance by ~10% through dataset refinement and targeted retraining, and I reduced training image collection time from 3 days to 1 by automating data acquisition. I additionally scaled a classification system from 20 to 100+ classes using Siamese networks and domain adaptation, and used Explainable AI plus K-Means clustering to address training data quality issues.

Data Scientist at EssilorLuxottica

October 1, 2021 - February 1, 2024

At EssilorLuxottica, I engineered and operationalized PyTorch-based computer vision models for manufacturing inspection and quality control. I developed an automatic eyewear aesthetic defect detector, expanded defect coverage through dataset analysis and cleaning, and built automated data processing pipelines to enable continuous retraining with reduced manual effort. I shipped a PyTorch segmentation model to estimate defect size, validated in real-world production, and also engineered a robotic keypoint detection model for eyewear temples. For deployment and scalability, I architected Azure cloud infrastructure to support inference for new temple designs.