Hello, I'm a student studying machine learning based in Cambridge, MA!

Bryan Gass

Profile Image

Experience:

Research Engineer

DEVCOM Research Laboratory - Boston, MA

Jun. 2023 - Current
  • Led research in extending context length for attention-free LLMs based on State Space Model (SSM) literature.
  • Conducted experiments on stabilizing dynamical systems to enhance SSM extensions and matrix parameterization.
  • Developed methods for efficient channel mixing.
  • Created sensitivity analysis metrics for comparing data influence on RNNs, SSMs, and Transformers weights.
  • Presented on specialized SSM matrices and future experiment directions at sister labs.
Machine Learning Intern

Dana-Farber Cancer Institute - Boston, MA

Aug. 2022 - Dec. 2022
  • Collaborated on applying state space models for training and up-scaling histopathology images in thesis project.
  • Thesis code and benchmarks available here
Research Intern

Nuance Communications - Burlington, MA

Jun. 2022 - Aug. 2022
  • Enhanced multi-GPU training for T5 models on medical documents; focused on attention mechanisms in large sequences.
  • Applied state space models and explored various attention mechanisms to extend context length in language models.
    • Worked toward applying state space models to language models
    • Researched attention mechanisms (linear-attention and self-attention) in an effort to finding methods toward increasing context length
Machine Learning Intern

Dana-Farber Cancer Institute - Boston, MA

Jan. 2022 - Jun. 2022
  • Developed and maintained PathML library, enhancing preprocessing, logging, and debugging; researched state space models for pathology image analysis.
  • Worked toward increasing resolution of patches used to process whole slide pathology images
    • Researched the application of state space models to aforementioned patches in an effort to leverage long context properties
Data Science Intern

Colgate-Palmolive - Piscataway, New Jersey

Jun. 2021 - Dec. 2021
  • Developed models optimizing supply chain strategy, cutting costs and emissions; automated problem detection alerts.
  • Created scalable preprocessing for data pipeline
  • Prepared, preprocessed, and managed large datasets
Research Engineering Intern

Nuance Communications - Burlington, MA

Jun. 2020 - Aug. 2020
  • Worked on the Automatic Speech Recognition (ASR) to aid in work within the mrecutil C++ library
IT Intern

Teledyne Marine Systems - Falmouth, MA

Jun. 2019 - Aug. 2019
  • Build and managed the platform of operations for various engineering projects
  • Assisted in server destruction, construction, subsequent upgrades and creating documentation to help future employees

Education:

M.S. in Computer Science, concentration in Machine Learning

Worcester Polytechnic Institute - Worcester, MA

Sep. 2022 - May. 2023
  • Thesis (In partnership with Dana-Farber Cancer Institute): Investigated differences in applying deep state space models for resolution agnostic training and zero-shot upscaling on large scale (gigapixel) histopathology images (mainly breast tissue).
B.S. in Computer Science

Worcester Polytechnic Institute - Worcester, MA

Sep. 2018 - May. 2021
  • Major Qualifying Project: Developed models, trainable on real world incompletely labeled data sets.
    • Research topics were built around positive unlabeled-learning disjunctive concepts in ILP and least general generalization
  • Interactive Qualifying Project: Worked with O.R.A.M. to develop and build an app that aids refugees
    • Features of the application were focused toward educating and protecting refugees and/or potential asylum seekers
A.S. in Mechanical Engineering

Cape Cod Community College - Barnstable, MA

Sep. 2016 - May. 2018
  • Worked as a tutor in the mathematics department
    • Aided students in topics: derivative, integral and multi-variable based calculus as well as linear algebra
© 2023 Bryan Gass. All Rights Reserved.