Ryan Lagasse

AI/ML Researcher & Engineer

Researcher Engineer @ Lockheed Martin
Director + Mentor @ Algoverse
Berkeley AI Fellow
Prev. Lecturer @ UConn

Active Research

Targeted LLM Steering: Mitigating Side Effects with Selective Feature Control

Improving activation steering vectors by formalizing constructive and destructive features to revert entangled features with negative side effects.

Accepted to RAISE Winter Expo

Extending Finetuning Scaling Laws for LLMs

Comprehensively selecting LLM sizes for small datasets based on token length and number of examples, improving on existing work.

Submitted to ICLR 2025 Workshop

Hybrid Quantum Algorithms for N-Body Simulations

Exploring hybrid quantum algorithms to accelerate n-body simulations with applications in astrophysics and molecular dynamics.

Accepted to QCNC 2025 (Oral Presentation)

RLOF: Scalable Fine-tuning for Small Datasets

Proposing a novel reinforcement learning-based optimization framework for training LLMs with limited data.

Targeting NeurIPS 2025

Persistent Refinement in Transformers

Improving chat bias benchmarks with innovative transformer steering methods for more robust and fair AI systems.

Targeting NeurIPS 2025

Additional Research Projects

Currently involved in 7+ ongoing research initiatives spanning mechanistic interpritability, alignment, control, and large-scale ML systems.

Various stages

Recent Highlights