The One About Reinforcement Learning
Reinforcement learning is being applied to train vastly different types of intelligent agents - from artificial neural networks to biological neurons. Join us for technical deep-dives into how RL methods are being used to create consistent AI characters and to teach human cells to navigate virtual environments.
More About the Sharings
Nasrudin Salim (CTO and Co-Founder, Sonzai Labs) will share on "Building AI-Native Characters That Stay in Character"
Nasrudin will share what it takes to build AI-native characters that feel coherent over time across multiple sessions. The real challenge is not generation, but state: how characters retain context, stay consistent, resist jailbreaks, preserve long-term memory, and remain believable while operating within the cost constraints of real B2C products. He'll walk through the problem space, the design tradeoffs, and practical solutions for making character systems reliable even on lower-cost models. This will cover personality layers, memory injection, semantic/context handling, and orchestration patterns that keep characters in character at scale. This infrastructure layer enables AI agents to function effectively as products. (Technical Level: 200)
Sean Cole (Independant Researcher) will share on "Teaching Human Neurons to Play Doom"
Sean will explore the ethics and technical implementation of using human cells in gaming environments, starting with existing approaches like manual encoding of behaviour in Pong. He'll dive into the specifics of their approach using REINFORCE encoders, stimulation patterns, decoders, and RL environments, explaining why Doom was chosen as the testing ground. The presentation will cover how this research can be applied to helping those with neurodegenerative diseases through RL models that teach brains how to learn, and close with how this technology is expanding in the Singapore context through collaborations with companies establishing major labs here. (Technical Level: 200)
*As the programme is still being finalized, do note that details are subject to change.
More About the Speakers
Nasrudin Salim is the CTO and Co-Founder of Sonzai Labs, a deep-tech AI startup focused on building the mind layer for AI agents. He previously served as VP of Engineering at Almanak, where he built orchestration and fund management AI agents for financial services. Earlier in his career, he was Lead AI Data Architect at OCBC and Head of Machine Learning Operationalization (MLOps) at DBS Bank.
Sean Cole did a polytechnic diploma in Cybersecurity at NYP, then went on to do a direct Master's at the University of Sussex's AI program. He now mostly works on independent research and has his own online marketplace startup
More About the Series
AI Wednesdays is Lorong AI’s weekly gathering, bringing together practitioners, researchers and innovators for technical discussions on research insights, product development and engineering practices.
Get involved: Learn more about Lorong AI | Speaker Sign-up | WhatsApp Community | LinkedIn | X