Alessandro Palmas
Senior Ai Ml Engineer @Ubisoft Montréal
Signup · Get unlimited contacts
WORK HISTORY
Senior Ai Ml Engineer @Ubisoft Montréal
Montreal, QC, CA
Develop and deploy advanced AI/ML systems, primarily based on Deep Reinforcement Learning, for integration into video game environments. Design algorithms focused on Curriculum Learning and Population-Based Multi-Agent RL (e.g, League Training) to create scalable and generalizable policies for autonomous agents. Lead the development of a distributed, modular RL training pipeline, enabling benchmarking of online and offline navigation algorithms at scale. Coordinate with internal research and engineering teams across France and Canada to align development with the company’s data-driven AI strategy. Partner with product teams to integrate AI models into proprietary game engines for use in automated testing, NPC behavior, and server-side simulation in live and in-development games. Collaborate with researchers and engineers to drive exploration of Imitation Learning, Hierarchical RL, Offline RL, and Continual Learning, while actively tracking extensions toward multimodal learning and LLM/VLM-enabled agents.
EDUCATION
Alta Scuola Politecnica
Multidisciplinary International Program
Politecnico di Torino
Master of Science (MSc)
Politecnico di Milano
Master of Science (MSc)
Politecnico di Torino
Bachelor of Science (BSc)
The University of Glasgow
Master of Science (MSc) by Research
SKILLS
ABOUT ALESSANDRO PALMAS
I\'m an applied AI/ML engineer with a passion for building intelligent agents, combining Reinforcement Learning, Multi-Modal Models, and Simulation to solve real-world problems and deploy robust, adaptive systems. Over the past 15 years, I’ve designed and implemented advanced solutions across a range of technical domains: Reinforcement Learning & AI Agents: Development of custom RL pipelines (online/offline), curriculum and league-based multi-agent training, imitation learning, and human-in-the-loop optimization. Hands-on with libraries like Ray RLlib, SB3, and custom frameworks. Large Language Models & Multimodal AI: Fine-tuning and alignment of transformer-based models via SFT and RLHF/RLAIF (PPO, GRPO and DPO). Exploring Visual-Language-Action Models (VLMs) and multimodal learning for grounded decision-making and control. Computer Vision & Perception: Deep learning models for detection, classification, semantic segmentation, behavior understanding, and synthetic data workflows, with applications in navigation, robotics, and real-time systems. Physics-Based Modeling & Simulation: Expertise in aerospace, defense, and environmental simulations: orbital mechanics, re-entry, parachute dynamics, CFD, and sensor modeling. Computational Geometry & 3D Environments: Mesh processing, ray tracing, collision detection, 2D/3D grid generation for simulation and rendering pipelines. I’ve built and contributed to multiple production-grade projects, including: DIAMBRA (https://diambra.ai): A competitive RL training platform for agents in video game environments (acquired in 2024) Artificial Twin (https://artificialtwin.com): My freelance consulting activity focused on delivering machine learning, physical simulation, and computational geometry solutions for international clients across aerospace, defense, cybersecurity, medical, manufacturing, and digital twin domains Currently expanding into LLMs, VLMs, their combination with RL and embodied learning systems, driven by the goal of bringing intelligent agents closer to real-world applications.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.