← Back to Portfolio
Personal · demo

Alakazam-VGC at AI Tinkerers Chicago

A talk on Alakazam-VGC — a fine-tuned LLM that answers competitive Pokémon battle-calc questions in natural language.

Alakazam-VGC demonstrates practical LLM fine-tuning and deployment for domain-specific applications, creating an AI assistant that processes natural language queries for competitive Pokémon VGC battle calculations. This portfolio project showcases end-to-end ML engineering from model customization through production deployment, addressing real user needs in the competitive gaming community.

Key Features

Multi-Intent Recognition: Processes three distinct query types relevant to competitive VGC players through fine-tuned language understanding • Domain-Specific Fine-tuning: Custom model training on Pokémon battle mechanics and calculation requirements • Production Deployment Architecture: Complete deployment pipeline demonstrating MLOps practices for LLM applications • Natural Language Interface: Eliminates need for manual calculations by accepting conversational queries about battle scenarios

Technical Implementation

The project addresses core challenges in LLM fine-tuning including data preparation, model selection, and deployment architecture decisions for specialized domains. The implementation demonstrates practical approaches to intent classification and response generation for gaming applications, with consideration for user experience and system scalability in production environments.

Skills & Technologies

Large Language Models Fine-tuning Natural Language Processing Model Deployment Python AI Application Development Intent Recognition
// stay in touch

Get Answering Machines in your inbox.

Occasional notes on AI, coding agents, and where this is all headed — plus the very occasional origami diagram.

Arjun Kirti Patel
Chicago, IL · folded with care
© 2026 Arjun Kirti Patel · MIT Licensed