Agentics Foundation: SLMs gotta get Fit
Agentics Foundation: SLMs Get Fit
Train Small Models with Reinforcement Learning
An interactive builder session for anyone curious about how AI models can learn through action, feedback, and rewards — not just prompts.
We will explore how 1B–3B Small Language Models can be connected to RL Gym-style environments using OpenEnv, trained to play old-school Atari-style games, and evaluated through real reward signals.
This is designed for both technical and non-technical participants. Developers can explore models, training loops, environments, and evaluation; non-technical builders can contribute through game strategy, reward design, testing, product thinking, and business use-case design.
The goal is not only to understand RL through games, but to see how the same concepts can improve real-world agentic systems. We will also share in-house demos showing how RL-style feedback loops can be applied to business workflows, operational decisions, customer experiences, automation, and agent performance.
📅 Schedule
Doors Open: 10:45 AM
Start Time: 11:00 AM
11:00 AM — Introduction: SLMs, RL, and Agent Behaviour
A simple explanation of how models learn through environments, actions, rewards, and feedback.
11:30 AM — OpenEnv + RL Gym Workshop
Connect a small model to a game environment, run the action loop, and understand observations, actions, rewards, and episodes.
12:15 PM — Build a Small Game Agent
Work in teams to connect a 1B–3B model to an Atari-style environment and test its behaviour.
1:15 PM — Fine-Tuning and Post-Training
Explore lightweight approaches to improving model behaviour through trajectories, rewards, and feedback.
2:15 PM — From Games to Business Workflows
See in-house demos of how RL principles can be applied to real business systems: decision-making, task optimisation, customer operations, workflow automation, and reliable agents.
3:00 PM — Demos, Learnings, and Wrap-Up
Teams share what they built, what worked, what failed, and where they see RL being useful next.
👥 Who Should Attend?
Developers, engineers, and AI builders
Product managers and designers
Startup founders and operators
Students and researchers
Open-source enthusiasts
Non-technical builders curious about AI agents and business workflows
No prior RL experience is required. Bring a laptop, curiosity, and a willingness to experiment.
Why Participate?
Learn how AI agents improve through feedback, not only prompting
Understand RL through a practical, visual game environment
Explore accessible 1B–3B open models
See how agent training concepts transfer to business workflows
Collaborate with both technical and non-technical builders
Leave with a clearer mental model of how reliable, reward-driven agents can be built
This is not a lecture-only event. It is a hands-on lab for builders who want to understand where agentic systems are going next.