Allen Roush
Foundation Model Architect & Director of Generative AI - Specialize in Generative Art, LLM, NLP, Argument Mining, and Model Explainability
Allen Roush is a Principal Machine Learning Architect with a specialization in Natural Language Processing (NLP), Argument Mining, and Model Explainability. He is an NLP researcher with a keen interest in taking Transfer Learning based search/summarization to the next level. Allen is curre
Show more
6.3
Years of Experience
Education
university of oregon
Companies
273 ventures, plai labs, oracle, intel, intel, oregon network research group, center for intercultural organizing
Reach out to Allen Roush via Email, InMail and SMS drip
by installing Chrome extension
Allen's contact details
Email (Verified)
gedXXXXXXXXXXXXXXXom
Github
HelXXXXXXXXXXXXle
Experience
2024 - Present
273 ventures
LLM Foundation Model Architect
- Training LLMs from scratch to achieve state of the art results, including results significantly outperforming GPT-3.5 - Working with the folks who showed that GPT3.5 can pass Bar exams.
2023 - 2023
plai labs
Director of Generative AI
- Running a team dedicated to creating and improving Generative AI products - Reporting to and working directly with most of the founders of MySpace and JamCity
2020 - 2023
oracle
Principal Machine Learning Architect - (OCI)
Doing all things involving Generative AI, NLP, Large Scale Language Modeling, and AI Evangelism at OCI - Did Enterprise Architecture, PoC demos, and delivered expert white glove technical and product support which directly landed new AI GPU customers running Generative AI and LLM workloads in the 10M YOY range on OCI. - Developed a comprehensive Data Science VM image for the OCI Compute Image marketplace, simplifying deployment for our users. - Created a dedicated Stable-Diffusion VM image for the OCI Compute Image marketplace to facilitate easy access to this powerful technology and near linear-performance distributed training/inference through support for Stable Horde - Published two papers at top NLP conferences and represented Oracle at these events. My most recent work, "Most Language Models Can Be Poets," explores the potential of constrained language models and was presented at a major AI conference hosted by Huggingface. All my papers are fully reproducible and accompanied by user-friendly demos to encourage further exploration and development. - Created, ran, managed, and administrated the Oracle Huggingface Page (https://huggingface.co/Oracle), ensuring that our organization remains at the forefront of the rapidly evolving field of AI and NLP technologies.
2018 - 2020
intel
Data Scientist
- Using NLP techniques to build recommender systems - Improving Model Explainability techniques for Black-Box ML models - Sourcing and training language models for company specific objectives.
2017 - 2017
intel
Software Engineer Intern
Interned with the Data Center Group (DCG) building infrastructure for HPC (High Performance Computing) products: - Designed a live monitoring dashboard for the continuous integration system, Jenkins. - Delivered and implemented a GUI for displaying Jenkins build and validation information, statistical usage of hardware resources, and a feature to generate comprehensive reports. - Implemented backend using Python, MongoDB, and Flask. - Frontend built with HTML, CSS, and JavaScript
2016 - 2016
oregon network research group
Programming Intern
- Wrote automation code to concurrently execute traceroutes through >1500 looking glass servers. - Used Java along with several API's and automation technologies such as Jsoup or Selenium. - Wrote parsing code to dump and properly format traceroutes into files within a folder for easy viewing - Implemented code used a factory design pattern - Added significant contributions into existing codebase
2013 - 2014
center for intercultural organizing
Intern
Worked gathering research for a non-profit organization related to equitable Transportation Infrastructure - Assisted senior policy analysts with research tasks - Frequently worked without supervision for long periods of time - Successfully completed various technical tasks for both offices, including time-sensitive tasks - Canvased for, helped to prepare, and lead events related to community organizing and outreach in the Beaverton office
Experience
63 Skills
Agile Methodologies
Agile Project Management
Algorithms
architecture
Artificial Intelligence (AI)
Automation
Bootstrap
C
C++
CentOS
Cloud Computing
Computer Vision
Cybersecurity
Data Mining
Data Science
Data Science
Data Scientist
Data Structures
Databases
Debate
Deep Learning
Docker
enterprise architecture
Flask
Generative AI
GPT
Gradio
HTML
Java
JavaScript
Jenkins
Kubernetes
language models
Large language models (LLM)
Leadership
Linux
Machine Learning
Machine Learning (ML)
Microsoft Excel
MongoDB
MySQL
Natural Language Processing
Natural Language Processing (NLP)
Natural Language Processing (NLP)
NoSQL
Object-Oriented Programming (OOP)
Project Management
Public Speaking
Python
Python (Programming Language)
R
Research Scientist
Scrum
Selenium
Senior Software Engineer
Software Testing
Software Validation
SQL
Streamlit
TensorFlow
Ubuntu
Virtualization
Windows
Education
2014 - 2018
university of oregon
Bachelor’s Degree
Computer Science
