​Agentics Foundation: SLMs Get Fit

​Train Small Models with Reinforcement Learning

​Most people meet AI through a prompt box.

​This session is about what happens when a small model can act, receive feedback, and improve.

​SLMs Get Fit is a highly hands-on builder session where we’ll explore how 1B–3B parameter Small Language Models can be connected to RL Gym-style environments using OpenEnv.

​We’ll use arcade-style games as a visual training ground for the core loop:

​observe → act → receive a reward → improve

​The point is not to build a game bot.

​The point is to understand how the same feedback loops can eventually power more reliable agents for real-world work.

​What we’ll explore

​Together, we’ll experiment with small models, game-like environments, rewards, trajectories, and evaluation. You’ll see how reinforcement learning can shape model behaviour beyond prompting alone—and why small, focused models can be surprisingly capable when trained for a specific task.

​We’ll also share selected enterprise use cases for SLMs, showing how these ideas can translate into workflows, operations, automation, document intelligence, customer experiences, and decision-making systems.

​And yes—there will be a few surprise demos. The goal is for you to leave thinking:

​“Wait… a small model can do that?”

​Hands-on means hands-on

​This is not a lecture where you sit back and watch.

​You’ll work alongside other builders, test ideas, see models interact with an environment, and understand how feedback changes behaviour over time.

​Technical participants can go deeper into models, environments, action loops, reward signals, and evaluation.

​Non-technical participants are equally welcome. You do not need prior RL experience or advanced coding skills. Our team will guide you through the session, and there will be meaningful ways to contribute through game strategy, reward design, testing, product thinking, and enterprise use-case design.

​Who should attend

​Builders, developers, engineers, founders, students, researchers, product people, operators, designers—and anyone curious about what comes after prompting.

​Bring a laptop, curiosity, and a willingness to experiment.

​Schedule

​Doors Open: 10:45 AM
Workshop: 11:00 AM – 3:00 PM

​11:00 AM — Small Models, Reinforcement Learning, and Agent Behaviour
11:30 AM — Hands-on OpenEnv + RL Gym Exploration
12:15 PM — Team Build and Experimentation
1:15 PM — Improving Model Behaviour Through Feedback
2:15 PM — Enterprise Use Cases, Surprise Demos, and Learnings
3:00 PM — Wrap-Up

​Come train, play, test, question, and build with us.


Attending in Bengaluru?
The main workshop is an in-person event in Bengaluru at the venue listed above.

​Joining from Assam or Guwahati?
A small in-person crew in Guwahati will join the Bengaluru workshop virtually. Connect with Mondweep Chakravorty for details.

​📍 Guwahati venue: O/o Chief Engineer (Water), PHED, Hengrabari
View on Google Maps