Experience

DirectAI

• United States
• Software Development
• 1 - 100 Employee

• Co-Founder

  • Aug 2022 - Present

  Vision models, without training data. Vision models, without training data.



Matician

• United States
• Computers and Electronics Manufacturing
• 1 - 100 Employee

• Deep Learning Engineer

  • Nov 2020 - Aug 2021

  Replaced all on-device convolutional neural networks with a unified framework based on a fused feature backbone and cutting-edge neural network architectures, greatly reducing total latency.Reduced object detection latency by 50% with no loss in accuracy, decreased depth estimation error by 30% with a slight speed increase, and enabled the addition of new subnetworks with minimal increases in latency, allowing for real-time simultaneous object detection, semantic segmentation, and depth estimation on edge hardware.Helped design novel approaches to neural anomaly detection, end-to-end neural pose estimation, and recurrent dense prediction.Extended previous work on self-supervision for temporally consistent depth estimations to provide a framework for others to use as a self-supervision signal for other dense prediction tasks.Worked as hiring manager for AI helping to build an in-house AI team, resulting in three hires. Show less

• AI Resident

  • Aug 2020 - Nov 2020

  Created and implemented novel self-supervised and semi-supervised training algorithms for stereo depth estimation neural network pipeline that created features consistent with the behaviors downstream usage of the network demanded, such as temporally consistent depth, smooth surfaces, well-behaved surface normals, estimation of per-pixel confidence, and sharper resolution of the edges of objects

• CV/ML Intern

  • Jun 2020 - Jul 2020

  Modified the production stereo depth estimation neural network to be compatible with our specific neural accelerator edge hardware with no loss in speed or accuracy.Created an alternative neural network architecture that was 30% faster with no loss in accuracy.



The National Institutes of Health

• United States
• Biotechnology Research
• 700 & Above Employee

• Civic Digital Fellow

  • Jun 2019 - Aug 2019

  Used supervised and unsupervised convolutional transformer neural networks to determine interesting features of healthy and cancerous cells as they migrate through 3D media via microscopy videos. Collaborated with microbiologists Dr. Andrew Doyle and Dr. Ken Yamada in the National Institute for Dental and Craniofacial Research's Cellular Biology Section. Used Microsoft Azure to demonstrate running machine learning tasks in the cloud on data uploaded via the NIH's STRIDES initiative in the NIDCR Office of Information Technology. Show less



Pearl Research, Inc.

• United States
• Technology, Information and Internet

• CTO

  • Jan 2018 - Feb 2019

  $10K in funding from Northeastern's Alpha Fund. Collaboratively built PCI-compliant AWS-based prototype cloud-based biometric authenticator with Python, Flask, C, MySQL, Nginx, and Apache. Designed and implemented novel fuzzy hashing neural network for biometric identification, achieving accuracy competitive with the state of the art for neural approaches. $10K in funding from Northeastern's Alpha Fund. Collaboratively built PCI-compliant AWS-based prototype cloud-based biometric authenticator with Python, Flask, C, MySQL, Nginx, and Apache. Designed and implemented novel fuzzy hashing neural network for biometric identification, achieving accuracy competitive with the state of the art for neural approaches.



Sol Wellness LLC

• United States
• Health, Wellness & Fitness

• CTO

  • Jan 2018 - Aug 2018

  Co-inventor on provisional patent application pertaining to automated depression risk stratification. Built HIPAA-compliant AWS-based backend to process audio and video and determine emotion of subject with Python, Flask, MySQL, Nginx, and Apache. Researched deep learning for depression stratification via video with Keras and Scikit-learn. Won Tsai CITY prize for the 2018 Yale Healthcare Hackathon, accepted into and participated in Yale Tsai CITY 2018 Spring Accelerator, Yale Innovation Summit, and Stanford MedX. Helped set up pilot studies at multiple medical schools including in Puerto Rico. Show less



Virtual Development Corp.

• Paid Intern

  • Jun 2016 - Aug 2018

  Collaboratively improved novel statistical machine learning, focusing on spam classification. Increased classifier speed by 10x using Numpy, Scipy, and Numba. Improved statistical stability using mpmath. Collaboratively improved novel statistical machine learning, focusing on spam classification. Increased classifier speed by 10x using Numpy, Scipy, and Numba. Improved statistical stability using mpmath.



Migdal Research LLC

• Paid Intern

  • Mar 2017 - Jul 2017

  Automated production collection of stock data with Python and Selenium, reducing labor costs Pair-programmed to create production stock symbol processing pipeline in Python Designed and implemented big data manipulation algorithms Performed original research using deep learning to predict the effect of news on stock and complement production model with Keras and Scikit-learn Automated production collection of stock data with Python and Selenium, reducing labor costs Pair-programmed to create production stock symbol processing pipeline in Python Designed and implemented big data manipulation algorithms Performed original research using deep learning to predict the effect of news on stock and complement production model with Keras and Scikit-learn



Bangor High School

• United States
• 1 - 100 Employee

• Student Researcher

  • Dec 2013 - Jun 2016

  Designed and simulated swarm-level cluster computing software for drones with IPython and OpenCV Designed and built drones with $3000 from Fast Track Grants via the Perloff Family Foundation Designed and simulated swarm-level cluster computing software for drones with IPython and OpenCV Designed and built drones with $3000 from Fast Track Grants via the Perloff Family Foundation