Experience

• 

  University of michigan

  Sept 2003 - Apr 2006

  Research assistant, medical imaging

  - Registered 3D breast ultrasound images taken at different times automatically with 90% accuracy for healthy patients and 73% accuracy for patients undergoing chemotherapy

• 

  Utopiacompression corporation

  May 2006 - May 2011

  Research and development engineer, computer vision

  - Co-authored grant proposals and successfully transitioned two projects from Phase I to Phase II securing government funding of $1.7 million- Designed a prototype of a scalable pattern-driven image compression tool in C++ and tested on datasets with 3 different resolutions achieving average compression ratio of 226- Developed an algorithm to extract building footprints in urban areas by information fusion from LiDAR data and high resolution aerial imagery with an average false alarm rate of 15% and accuracy of 84% Show less

• 

  Cgg

  Jan 2012 - Jul 2014

  Software developer, seismic imaging

  - Optimized a Kirchhoff Depth Migration module processing 100TB of seismic data on GPU clusters using C++ to get results up to 15x faster- Identified erroneous or missing data during data collection by performing quality check on raw maritime seismic images improving accuracy of data by 99%- Added new features to the QT based proprietary 2D seismic image GUI to improve visualization of data- Collaborated with a global team of researchers, software developers and end-users

• 

  Metis

  Jan 2018 - Apr 2018

  Data science fellow

  Metis is a competitive 12 week immersive Data Science training program covering topics such as Statistics, Machine Learning, Programming, Communication and Design.Using Python as the backbone, I designed and built the following end-to-end projects using real-world data that yield critical insights1) Meetup Recommender System for Organizers - collected data using Meetup API and built a Content-Based Recommender System for Meetup group organizers to help find synergies in similar groups, hence maximizing the impact they can make on their community2) Topic Modeling on Dietary Supplement Labels - discovered “topics” that group dietary supplements with similar safety concerns by applying Latent Dirichlet Allocation on the text information printed on the labels3) Predicting Flight Delays in United States - proposed a flight delay prediction model which passengers can use a few months before their trip all the way to the day of the trip with the best F2 score of 0.75 on unseen data 4) Predicting Campaign Funding on Kickstarter - scraped and cleaned Kickstarter project data to predict the amount of funding a campaign will get based on the information the creator puts on their project page with an R-squared of 0.74 on unseen data5) Billboard Placement Strategy - analyzed New York City subway station entry/exit data to draw insights about the neighborhoods and times which have the highest traffic density for strategic placement of billboards for a food delivery company6) Austin Displacement Alert - Part of a team to build a heat map visualization tool of displacement risk in Austin, TX. Created a framework for the automatic classication of city blocks as at-risk of eviction or not. Cleaned the feature datasets to bring to the same geographic resolution. Built a baseline Logistic Regression binary classier using historical eviction data.Tools used: Python, SQL and NoSQL databases, AWS, Machine Learning algorithms and visualization libraries, GIT Show less

• 

  Cisco

  Nov 2018 - Sept 2022

  Data scientist

  Delivered strategic and actionable insights through data analysis (tools: Python, PostgreSQL) and visualization (tools: Power BI, Dash) to the Cisco DevNet team to drive business growth, improve user experience and brand loyalty- Performed cohort analysis to understand user retention and create customized personas for DevNet products. - Increased accessibility and usability of usage data to support DevNet growth* Standardized data and automated data mining process resulting in a 90% decrease in the time needed to gather insights from user data. This strengthened marketing strategies resulting in growing the user community by 75%* Formulated key metrics communicating the state of business, published daily in a Webex space and viewed by the entire DevNet team* Architect for >25 PowerBI visualizations to support both operational metrics and executive review- Partnered with product, design, engineering, and marketing teams to design data pipeline and formulate metrics for go to market strategy- Collaborated with various business units within Cisco to formulate key product metrics and identify synergies with the DevNet community- Led cross-functional objectives and key results (OKR) for DevNet to formulate key product metrics that track revenue growth Show less

• 

  Roku inc.

  Sept 2022 - now

  Senior data scientist