Tommaso Colombo

Tommaso is the Head of Artificial Intellingence at Spindox S.p.A. since 2021. He leads the R&D and industrial delivery in the AI domain, managing a group of 50+ professionals.
Tommaso holds a PhD in Operations Research applied to Machine Learning, through which he researched new training algorithms for Deep Neural Networks and Reinforcement Learning, also thanks to its visiting research period at Northwestern University, Chicago.


Your job title

Head of AI

Your company

Spindox


Session

12-11
11:30
40min
How LLMs can help sw development, math-modeling and forecasting
Tommaso Colombo, Alessandro Pinzuti, Nicolò Mazzi

Spindox has been developing its Gen. AI framework, named Atlante, for over 2 years. Atlante integrates key open-source libraries like LangChain, Llama-Index, and LangGraph, among others. Furthermore, it employs best-of-breed LLMs – both commercial and open-source ones – and relies on advanced strategies for chunking, RLHF and multi-agent components (i.e. agentic pattern). It is designed for supporting in a range of applications spanning from math-optimization algorithms development to no-code simulation models building, from support in the whole software life-cycle to advanced RAG. Its modular, cloud-agnostic architecture ensures flexibility and cost-effectiveness. Enhanced by caching and knowledge graphs, Atlante optimizes processing, reduces costs, and improves accuracy. It is widely adopted across industries for various applications, from AITSM to coding, test automation and marketing documents generation. One of the hands-on deep-dives will be the application to automatic math-optimization solution algorithms development, which was recognized the first price at ICML 2024 – Challenges on Automated Math Reasoning.
https://www.codabench.org/competitions/2438/
https://medium.com/@alessandropinzuti

Auditorium