Sam Anzaroot pdf version
Mobile: (718) 207-2887 Email: email@example.com Website: samanz.com Github: samanz LinkedIn: sam-anzaroot Google Scholar: Sam Anzaroot
Feb. 2021 - Present Applied AI Researcher
- First employee of the company. Built and launched the company's first product. Ran meetings with potential external clients and managed final deliverables.
- Had additional role of project manager, running weekly project management meetings and managing assigned issues for entire engineering team.
- Built core machine learning/AI/NLP technologies using state-of-the-art transformer language models, such as ROBERTA and T5.
- Created automatic training, evaluation, and deployment pipelines using Kubeflow and NVIDIA Triton Inference Server.
- Sped up large language models using model distillation and NVIDIA FasterTransformer for an around 10x-20x speed up.
- Managed AI team employees and mentored interns.
July 2019 - Sept. 2020 Principal Data Scientist
Jan. 2017 Senior Data Scientist
Feb. 2015 Data Scientist
Sept. 2014 Software Engineer in Data
- Helped grow the AI team over six years by leading multiple high-profile projects, advocating internally for state-of-the-art techniques, leading an AI reading group and mentoring interns. Communicated with stakeholders including product managers, HCI researchers, designers, domain experts, and engineers.
- Led and contributed to team focused on automatically generating summaries of public safety events detected from social media posts. The team utilized seq2seq LSTM and Transformer deep-learning models, and ran a user study and deployed a human-in-the-loop system for summary writing to production which sped up summary writing by 2x.
- Led and contributed to geo-prediction team, focused on detecting mentions of locations in unstructured text and geocoding mentions to points on earth. Trained and deployed a neural network conditional random field model and neural network LambdaRank model, drastically increasing location precision on Dataminr content.
- Led and contributed to automation team, combining multiple different models in a pipeline for full content automation. This project resulted in the full automation of the majority of content sent by Dataminr.
- Worked as IC on various projects, including a novel language-identification model for social media, a text-based topic prediction model, a novel neural-network library built in Scala, a named entity recognition model for social media, and a label annotation platform.
Oracle Labs - East
Feb. 2014 - June 2014 Research Intern
- Researched methods for highly parallel probabilistic inference on conditional random fields (CRFs) using GPUs.
- Created a GPU version of the belief propagation algorithm written in CUDA. Optimized this implementation to allow for 200x speedup in inference and 100x speedup in training of CRFs over sequential implementation.
IESL, University of Massachusetts — Amherst
Sept. 2011 - Feb. 2014 Research Assistant
Advisor: Andrew McCallum
- Performed NLP and ML research focusing on undirected graphical models and information extraction.
- Oversaw creation of a novel citation extraction dataset, the largest and most fine-grained openly available dataset for this task.
- Developed method for more robust inference in conditional random fields using extensions to Lagrange relaxation methods called soft dual-decomposition with applications in citation extraction, retrieving new state-of-the-art results on the citation extraction task.
March 2016 - Sept. 2016 Data Science Volunteer
- Implemented methods for automatically extracting metadata from research documents to assist researchers in performing systematic literature reviews.
- Helped build and deploy a machine learning enabled systematic review web application currently in use by researchers available at colandrapp.com
University of Massachusetts — Amherst
Feb. 2014 MS in Computer Science
3.6 / 4.0
Graduate level coursework
- Machine Learning
- Statistical Inference I
- Automated Knowledge Based Construction
- Graphical Models
- Research Methods
- Advanced Databases
- Distributed Operating Systems
- Artificial Intelligence
Queens College — City College of New York
June 2011 BS in Computer Science
3.75 / 4.0
Magna cum laude
Graduate level coursework
- Natural Language Processing
- Machine Learning
- Human Computer Interaction
Unsupervised Detection of Sub-Events in Large Scale Disasters.
AAAI Conference on Artificial Intelligence, 2020.
Crisis Sub-Events on Social Media: A Case Study of Wildfires
AI for Social Good Workshop at the 36th International Conference on Machine Learning (AISG@ICML 2019), 2019.
Using machine learning to advance synthesis and use of conservation and environmental evidence.
Conservation Biology, 2018.
Learning Soft Linear Constraints with Application to Citation Field Extraction.
52nd Annual Meeting of the Association for Computational Linguistics (ACL2014), 2014.
A New Dataset for Fine-Grained Citation Field Extraction
ICML Workshop on Peer Reviewing and Publishing Models (PEER), 2013.
Joint Inference for Crossdocument Information Extraction
20th ACM Conference on Information and Knowledge Management (CIKM2011), 2011.
Machine Learning Frameworks
- Postgresl - SQL
- Machine learning
- Deep learning
- Natural language processing