Open to opportunities

JB Bryant

@jbbryant

Message

I build and evaluate LLM agents, reinforcement-learning systems, and optimization pipelines for complex AI challenges.

China

Message

What I'm looking for

I'm looking to build and evaluate LLM agents, reinforcement-learning systems, and optimization workflows where rigorous experimentation, verification, and human control of system design matter.

I've built competition-grade LLM agent and optimization systems on Kaggle, including a parallel multi-agent system for 400 verifiable ONNX tasks that earned a solo Silver Medal in the 2026 NeuroGolf Championship.

For the Pokémon TCG AI Battle Challenge, I designed a relation-aware Transformer policy, built behavior cloning from official replays, and developed PPO training over complete games. The system achieved over 80% teacher-forced exact action accuracy, approximately 70% head-to-head win rate against a 17-opponent league, and peaked in the public leaderboard Top 5.

I'm focused on evaluation design, model behavior, reinforcement learning, failure diagnosis, and optimization. I also placed first in the MetaX x ModelScope challenge by fine-tuning Qwen3-1.7B with LoRA and GRPO, improving MATH-500 mean accuracy from 0.7367 to 0.7776.

Experience

Work history, roles, and key accomplishments

Applied AI Practitioner

Kaggle

Applied AI practitioner focused on LLM agents, evaluation, model behavior, reinforcement learning, and optimization. Kaggle Competitions Expert with medal finishes across agent simulation, optimization, and LLM-oriented competitions, plus a first-place finish in a small-model GRPO post-training challenge.