MSc Dec 2026 · Open to AI engineering roles

Francesco Pio De Dominicis

Local models.
Real impact.

AI engineer and MSc candidate at Politecnico di Torino. Passionate about local inference and projects that truly matter.

See my work Download CV
Dec 2026MSc graduation · PoliTo
10+projects on GitHub
Agents → Edgefrom LangGraph to MLX

Featured · MSc Thesis

Is this answer actually supported by its sources?

A specialist decision model that verifies, sentence by sentence and in a single forward pass, whether a RAG answer is supported by its sources. Fine-tuned Qwen2.5 (0.5B / 3B) with LoRA on MLX, running fully on Apple Silicon, with a native SwiftUI app.

  • Qwen2.5
  • LoRA
  • MLX
  • RAG
  • SwiftUI
View on GitHub ↗
groundone · verify2 / 3 supported
GroundOne fine-tunes Qwen2.5 with LoRA.✓ Supported
It runs fully on-device on Apple Silicon.✓ Supported
It was trained on ten billion documents.✕ Unsupported
Illustrative example

Selected work

Things I’ve built.

All repositories ↗
macOS · On-device AI

Tempra

A menu bar monitor for your Mac’s vital signs, with on-device forecasts powered by Mistral Timestral.

  • Swift
  • SwiftUI
  • Forecasting
LLM Agent

Travel Assistant

Conversational trip planner where the LLM understands and explains, while the code decides and guarantees.

  • Python
  • LLM
  • Tool use
Explainable AI

Hierarchical CBMs

Coarse-to-fine Concept Bottleneck Model with LLM-generated concepts and superclass gating on CIFAR-100.

  • PyTorch
  • XAI
  • LLM concepts
Explainable AI

Interpretable SER

Speech emotion recognition that explains its predictions through Concept Bottleneck Models.

  • PyTorch
  • Audio
  • XAI
Computer Vision

Road Anomaly Segmentation

Open-world segmentation for autonomous driving with Mask2Former (EoMT) and Outlier Exposure.

  • PyTorch
  • Transformers
  • OOD
Systems · Rust

Ruggine

Stateful chat built on Tokio: gRPC streaming over HTTP/2, PostgreSQL history, Sled cache, TLS 1.3 and Argon2.

  • Rust
  • gRPC
  • PostgreSQL

Toolkit

Tools I reach for.

AI Engineering
  • LangGraph
  • LangChain
  • RAG pipelines
  • Hybrid search (BM25 + dense)
  • PGVector
  • FAISS
  • Chroma
  • Azure OpenAI
  • Langtrace
  • Langfuse
Local inference & fine‑tuning
  • MLX
  • llama.cpp / GGUF
  • Ollama
  • LM Studio
  • Hugging Face
  • LoRA fine-tuning
  • Unsloth
Machine Learning
  • PyTorch
  • TensorFlow
  • scikit-learn
  • pandas
  • NumPy
  • Computer Vision
  • Explainable AI
Languages
  • Python
  • Rust
  • Swift
  • Kotlin
  • Java
  • C
  • TypeScript
  • JavaScript
  • Dart
  • SQL
Backend
  • FastAPI
  • Node.js
  • Express
  • Spring Boot
  • Apache Kafka
  • gRPC
Data
  • PostgreSQL
  • MySQL
  • SQLite
  • MongoDB
  • Redis
  • Apache Spark
  • Apache Hadoop
Apps & Frontend
  • SwiftUI
  • Flutter
  • React
  • HTML
  • CSS
DevOps & Cloud
  • Docker
  • Kubernetes
  • Helm
  • Argo CD
  • GitHub Actions
  • Terraform
  • Azure
  • Prometheus
  • Grafana
  • OpenTelemetry

Experience

Where I’ve worked.

  1. Mar 2026 — Present

    Generative AI Engineer · Intern ● Now

    Accenture · Turin

    Multi-agent workflows in LangGraph (supervisor / router patterns, tool calling) and RAG over enterprise documentation, deployed on Azure OpenAI with evaluation and guardrails.

  2. Jun — Aug 2024

    Software Engineer · Intern

    Youbiquo · Cava de’ Tirreni

    Flutter app integrated with OpenRMF for real-time teleoperation, telemetry and command dispatch of heterogeneous robot fleets.

  3. 2017 — Present

    Football Referee

    AIA · FIGC

    Split-second decisions under pressure, clear communication with assistant referees.

Education

  1. 2024 — Dec 2026

    Politecnico di Torino

    MSc Computer Engineering · AI & Data Analysis

    Thesis: GroundOne — sentence-level grounding verification for RAG.

  2. 2021 — 2024

    Università degli Studi di Salerno

    BSc Computer Engineering · 109/110

    Thesis: ROAM — cross-platform control of heterogeneous robot fleets.

Let’s build something
grounded.

Open to AI engineering roles · Based in Turin, Italy

fpdedomi@gmail.com LinkedIn GitHub Download CV ↓