> vision · llm · edge · deploy

Riccardo Gasparini

Freelance AI engineer. I put vision models and LLMs inside products and factories, from prototype to deployment.

Email meLinkedIn

01

What I do

LLMs, agents and RAG

Assistants that answer over company documents, structured extraction from PDFs and scans, agents that query databases and systems. Cloud or self-hosted GPU models, with measurable evaluation before production.

  • RAG
  • LangGraph
  • Claude / OpenAI
  • vLLM
  • Docling
  • Qdrant

Industrial computer vision

Inspection, anomaly detection, people and vehicle tracking, shop-floor safety. From camera selection to the in-line model, with hands-on experience on industrial cameras and synchronized acquisition.

  • PyTorch
  • OpenCV
  • FLIR / Allied Vision
  • GigE Vision
  • YOLO
  • Tracking

Data, edge and deployment

Forecasting on IoT time series, satellite and geospatial data, models on embedded hardware. APIs, containers and CI/CD pipelines so the model actually reaches the people who use it.

  • FastAPI
  • Docker / Kubernetes
  • ONNX
  • CUDA
  • Hailo / GAP8
  • PostGIS

02

Work

A selection of recent projects, no names: what matters is the problem and how I solved it.

Industrial group

Multi-company RAG platform over internal documents

Problem
Thousands of technical documents and procedures scattered across group companies, impossible to query.
What I did
Docling ingestion, hybrid dense and sparse search with reranking, visual page retrieval, streaming answers with per-company permissions. Continuous evaluation with RAGAS, Kubernetes deployment.
  • Docling
  • Qdrant
  • BGE-M3
  • vLLM
  • FastAPI
  • Helm

Medtech regulatory

LLM-driven medical device classification

Problem
Classifying a device under the EU MDR means walking a long, ambiguous decision tree.
What I did
A 146-state machine follows Annex VIII while an LLM answers each question from the technical file. RAG over guidance documents, serverless backend with streaming, React frontend.
  • Claude
  • AWS Lambda
  • OpenSearch
  • Bedrock
  • React

Technical documentation

OCR and structured extraction from product sheets

Problem
Technical parameters locked in PDFs, scans and Office files of every format, to be mapped to a schema.
What I did
Service built on a LangGraph graph: filter, extract, validate. Docling parsing, OCR and self-hosted vision-language models, output validated with Pydantic schemas. Multi-tenant, CI/CD, Kubernetes.
  • LangGraph
  • Docling
  • VLM
  • Pydantic
  • vLLM
  • GitLab CI

Manufacturing · EU H2020 project

Shop-floor safety with computer vision

Problem
Detect falls, entries into danger zones and safety distances between workers from existing cameras.
What I did
Pipeline with instance segmentation, pose estimation, tracking and a fall classifier, face anonymization. Events published to a FIWARE broker, dashboard for the safety manager, deployed on the project Kubernetes cluster.
  • PyTorch
  • YOLACT
  • Pose
  • FIWARE
  • Streamlit
  • ArgoCD

Precision agriculture

Soil forecasting and irrigation advice from IoT and satellite

Problem
Field sensors, weather forecasts and Sentinel-2 imagery to turn into a decision: how much to irrigate and fertilize, and when.
What I did
Forecasting models (Random Forest and Temporal Fusion Transformer) on soil water potential, agronomic recommendation engine, NDVI vigor maps with prescription zones exported for machinery.
  • scikit-learn
  • PyTorch Forecasting
  • ONNX
  • Sentinel-2
  • InfluxDB
  • FastAPI

Railway · research

Anomaly detection for line inspection

Problem
Find defects and anomalous objects on insulators and line components from thermal and RGB images, on embedded hardware.
What I did
Autoencoder-based anomaly detection, localization and classification with a convolutional network, optimized for on-board devices. Published at ICPR 2020.
  • Autoencoder
  • CNN
  • Thermal
  • Embedded
  • ICPR 2020

03

Path

Portrait of Riccardo Gasparini
modena · 44.65N 10.93E
10+
years in AI and vision
5
publications
1
H2020 project
100+
teaching hours
Eng.
licensed engineer

  1. 2022 · today

    Freelance AI engineer

    Consulting and development for manufacturing, agritech, medtech and service companies: RAG, agents, vision, forecasting, deployment.

  2. 2019 · 2024

    Co-founder and CIO of an AI startup

    Technical lead: architecture, team, product. EU H2020 project as technology provider.

  3. 2016 · 2022

    Research fellow, AImagelab UNIMORE

    Computer vision and machine learning: people tracking, railway inspection, embedded vision, synthetic data. Five international publications.

  4. 2017 · 2020

    Teaching

    GPU programming with CUDA (second-level master), machine learning and video analytics, fundamentals of computer science. Five theses supervised.

  5. 2016

    MSc in Computer Engineering

    UNIMORE, 108/110. Thesis on real-time 2D-3D matching. Licensed engineer, Modena register.

  6. 2012

    BSc in Computer Science

    UNIMORE. Thesis on image reconstruction on NVIDIA GPUs.

04

Stack

Models and ML

  • PyTorch
  • ONNX
  • TFLite
  • CUDA
  • scikit-learn
  • Optuna

LLM

  • Claude
  • OpenAI
  • vLLM
  • Ollama
  • LangChain / LangGraph
  • Docling
  • Qdrant
  • OpenSearch
  • RAGAS

Vision

  • OpenCV
  • YOLO
  • YOLACT
  • FLIR Spinnaker
  • Allied Vision Vimba
  • GigE / PTP
  • Hailo
  • GAP8

Backend and deployment

  • FastAPI
  • Django
  • Docker
  • Kubernetes
  • Helm
  • GitLab CI
  • AWS
  • Keycloak
  • PostGIS
  • InfluxDB

Frontend and mobile

  • React
  • Astro
  • Flutter

05

Publications

  1. 2021ICCVMOTSynth: How Can Synthetic Data Help Pedestrian Detection and Tracking?
  2. 2020ICPRAnomaly Detection, Localization and Classification for Railway Inspection
  3. 2020AI4RAILSAnomaly Detection for Vision-based Railway Inspection
  4. 2017IEEE IVEmbedded Recurrent Network for Head Pose Estimation in Car
  5. 2016AVROptimizing Image Registration for Interactive Applications

06

Contact

Have a problem that looks like one of these? Send me a couple of lines: I reply within one working day.

Modena, Italy. Remote, or on site in Emilia-Romagna.