I currently lead the match pod at CMB, which is a cross functional data science and engineering team that supports our matching algorithm. The analytics functional also reports to me.* Led initiatives that increased current user retention to all time highs via a combination of modeling and business logic* Worked with data scientists to produce novel analyses to inform strategic decisions for the future of the product and user acquisition strategies* Worked with data scientists and engineers to plan and execute on an overhaul of our backend matching platform / service* Drove new strategic product initiatives including "contextual recommendations" and a "deeper profile"

* Built the data science team from scratch by hiring 3 data scientists by the end of 2018* Led the match pod, a hybrid data science and engineering team dedicated to supporting our matching algorithm* Worked cross functionally to support the revenue and growth pods via experimental design, experiment readouts, and analyses to drive more insight into user behavior

* Architected and implemented hybrid offline / realtime scalable collaborative filtering system for recommendations via generalized matrix factorization pipeline (SVD) that increased revenue by 8%* Migrated from Cassandra -> Redis for our recommendation datastore, which reduced engineer and DevOps time spent managing our recommendation infrastructure by an order of magnitude

* Built A/B testing framework for our matching algorithm to enable deployment of new experiments with simple configuration file changes and minimal backend implementation* Implemented a massive overhaul to our matching algorithm via realtime heuristic models in combination with logistic model trained offline to solve supply / demand problem and maximize coverage of female users being liked. Increased unique female users connecting per day by 90%* Increased effective throughput of our recommendation pipeline for daily matches by 200% through various engineering optimizations (e.g. using geo hashing for initial pool of candidate matches)* Built logistic regression model for female users that increased like rate by 25%

* Implemented 2 major data model / database migrations, most notably from Titan -> Cassandra for our curated recommendations and from Postgres -> Cassandra for user photos* Decreased latency of live match generation in web requests by 80%, and increased general stability of our recommendation infrastructure* Redesigned in-house priority queue (Redis) to support granular prioritization, scheduled tasks per user, and retry mechanisms (adapted for specific business needs that couldn't be supported by Amazon SQS or Celery)

I worked as a hybrid data analyst / consultant to help venture capital LPs and GPs benchmark returns.* Owned the software implementation of our index of returns for venture-backed companies (similar to S&P 500 but for venture)* Built logistic regression model to predict likelihood of a venture-backed company going out of business since its last round of funding* Published quarterly report for our subscribers, including quantitative returns, and qualitative topics of interest* Managed teams of researchers to build our dataset of funding rounds and exit events (IPOs, acquisitions, failures), in addition to our vendor supplied datasets* Consulted with LPs to benchmark aggregate returns of venture portfolios, and performance of individual funds* Consulted with GPs to help lend data to support their fundraising efforts and validate their investment strategy* Developed NPV model to benchmark custom venture segments (e.g. healthcare, early stage) against equivalent public market investment strategy* Implemented variations of our venture index to take into account lockout periods after IPO by joining with public market datasets