Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Pokemon LLM Battle Calculations
Learn how to build a natural language assistant for Pokemon battle calculations, five times faster than existing tools. This talk covers LLM fine-tuning, deployment, and practical project scoping.
I’ll present my personal project, which involves building a command line-like assistant for competitive Pokemon players. I built a web app that accepts natural language queries, and returns relevant battle calculations for Pokemon players about 5x faster than the current best calculation software. Very often when playing online, players will need to estimate the probability that a Pokemon will knockout, two-hit-ko, or outspeed the opposing pokémons with incomplete information about the game state. Online calculators offer this capability, but you have to navigate airplane dashboard-like interfaces to use them. Since players already describe these game states in a specific lingo, I decided to create an app that can directly translate this natural language query to the calculation!
Finetuned BLOOM LLM parses natural language Pokemon VGC speed and optimal EV calculations.