I have been acting as project manager and technical lead for a team developing a full computer vision system for a real world application. On the management side this includes managing a team of 3-6 engineers, managing the project github repository, doing regular code reviews to ensure high quality and robust code is being implemented, developing gantt charts to communicate and track project timelines, allocating resources across a diverse team, providing weekly updates to key project stakeholders, managing a kanban board to track individual project components, and developing project workflow diagrams to clearly illustrate how the different tasks fit together. On the technical side, this includes building custom cameras from individual hardware components, developing automatic camera calibration systems, implementing sensor fusion between multiple sensor systems, developing a multi-object tracking and keypoint detection system, and designing custom metrics based on the MOT output.

• Developed and implemented machine learning models to help detect fraud and overpayments• Used clustering (intelligent vectorization and k-means) to develop member profiles

● Lead Data Science team to develop machine learning based solutions to clinical problems. ● Work with physicians to develop grant proposals and write scientific papers.● Certified to retrieve data from Epic's EHR, clarity ambulatory certification specifically.● Developed machine learning model to pre-operatively predict highly significant surgical outcomes of total joint replacements with >90% accuracy. Used this model to help identify effective interventions at the individual patient level.● Developed series of deep learning models to use hand x-rays to identify a patients likelihood of osteoporosis. Achieves >80% accuracy. This is not something doctor's or radiologists are trained to look for in hand x-rays. Uses both deep learning based segmentation and classification.● Developed gradient-boosted decision forest based model to identify the likelihood of a patient with any type of cancer dying within 6 months of diagnosis. Works with >85% accuracy.● Examined data on how doctors write notes in the Electronic Medical Record to help identify key factors to optimize system usage and save doctors times.● Developed computer vision based method to map lymph system from infrared videos of the human hand.● Developed natrual language processing models to help classify patients by diagnosis from free text doctors notes in EHR.● Developing 3D deep learning segmentation model to segment volumes of scar tissue from full 3D ultrasound volumes of tendon injuries.● Developing recurrent neural network to identify which alarms in the pediatric ICU are actionable and which are not● Developing machine learning model to analyze patient's speech patterns to determine disease severity of cognitively impaired patients● Developing machine learning model to help asses primary care doctors ability to interact with suicidal patients via audio and visual analysis of mock patient interviews● Developing deep learning model to segment ulnar nerve from ultrasound images

Adjunct Instructor for the Goergen Institute of Data Science at the University of Rochester.Teaching DSC 481, Apllications of AI and Deep Learning to Healthcare.This is a graduate level course I designed myself to teach students how to take machine learning and deep learning theory and directly apply it to healthcare data. It includes teaching ML and DL theory, along with specifics of how to deal with medical data, and real world examples of how I have applied these ML and DL techniques in medicine.

● Developed software to download a variety of information from the twitter API such as tweets from specific geographic regions, the social graph of a network of over 400,000 users in the Rochester area, and the tracking of tweets from specific individuals and businesses.● Maintained all data sets gathered from the twitter API since summer of 2014 (several TBs).● Developed various filtering methods to create new data sets from the data already collected from the API.● Assisted in using twitter to identify, in real time, restaurants that have had a food poisoning outbreak in Las Vegas.● Assisted in using twitter to identify alcoholics/alcoholism in the Rochester area.● Assisted in using both heuristic and machine learning methods to identify age and LGBT orientation in Rochester.● Worked in a large and varied team consisting of individuals from several different fields.

● Singlehandedly set up a data capture infrastructure to continuously capture and store a large variety of twitter data (responsible for the capture of about 1TB of data per month).● Designed website to display in real time the status of the data capture.● Automatically tracked and analyzed anomalous peaks in twitter data related to several specific subjects.● Used OpenCV to analyze usage of images in twitter by language community (specifically images from phones).● Developed tools to use for sentiment analysis on specific tweets.