Nov 2018 – Dec 2020 · 2 yrs 1 mo

• At Clarity (clarity.ai) we are solving the problem of inefficient and unequal allocation of capital by developing a universal framework and an automated tool that can be readily applied to rate companies, projects, governments and NGOs.

Sep 2016 – Nov 2018 · 2 yrs 2 mos

Sep 2017 – Nov 2018 · 1 yr 2 mos · Los Gatos

Jan 2017 – Sep 2017 · 8 mos · San Francisco

Apr 2015 – Jan 2017 · 1 yr 9 mos · San Francisco Bay Area

• Miroculus is a life science company that has developed an accurate, easy to use, non-invasive and affordable microRNA detection platform that can radically improve the ability of life science research and healthcare providers to diagnose, treat, and monitor diseases at a molecular level through the examination of microRNAs.

Jan 2014 – Feb 2015 · 1 yr 1 mo · Los Gatos, CA

• Managing a talented research team imagining the next version of our world-class recommender algorithms. These algorithms drive most of the Netflix experience, including personalization of the home page, search recommendations, Top picks row, Because you watched row, and game-based recommendations.

Sep 2013 – Jan 2014 · 4 mos · Los Gatos, CA

• Data-driven muti-touch attribution modeling
• Bottoms-up lifetime value (LTV) modeling
• Content-based, LTV-aware modeling of FB audiences

Sep 2011 – Sep 2013 · 2 yrs · Los Gatos, CA

• Recommendation algorithms and predictive analytics.
• Social cold-start of recommendation algorithms: Developed, tested, and implemented a ranker that predicts which friends of Facebook-connected new subscribers will be closer on movie taste similarity. The model uses gradient boosted decision trees on a variety of social- and content-based features.
• Adaptive row selection: Developed, tested, and prototyped a scoring metric for a "recommender of recommendation algorithms" based on past performance. The model uses logistic regression with elastic net penalty, and k-means clustering on a variety of user-, content-, and algorithm-based features.
• Markov-chain based movie similarity: Explored, derived, and prototyped different data-driven metrics of movie similarity based on users' play data to improve movie recommendations. Explored the breakdown of user population by usage patterns using Expectation-Maximization.
• Short-term attrition modeling: Developed a parsimonious Bayesian model for next-day attrition based on historical play rates.

Oct 2010 – Sep 2011 · 11 mos · New York City, NY

• Price verification for derivative instruments
• Implemented calibration routines for stochastic pricing models of commodity derivatives, allowing for automated price verification. Developed tools to analyze the impact of resulting price differences across the firm’s portfolio.

Jan 2007 – Oct 2010 · 3 yrs 9 mos · Cambridge, MA

• State-space modeling of combined EEG+MEG time series (PhD Thesis):
• Designed efficient algorithm for time series modeling and estimation. Hybrid distributed/shared memory implementation for large data sets in a Linux clus- ter using C++, MPI, ScaLAPACK, and OpenMP.

Jan 2006 – Jan 2007 · 1 yr · Cambridge, MA

• Integration of EEG and MEG data in minimum 2-norm estimation:
• Characterized spatial resolution and localization accuracy of inverse operator using a large clinical dataset. Designed relevant summary metrics. Results were published in peer-reviewed, high impact-factor international journal.

Jan 2005 – Jan 2006 · 1 yr · Cambridge, MA

• Data analysis of high-speed two-photon imaging:
• Developed a new interactive analysis tool to monitor temporal profiles of cell activity in real-time, extracted from large data sets, using Matlab and C++.
• Statistical modeling of neural activity: Designed new way of characterizing and displaying single neuron spiking activity using discrete stochastic processes theory.

Jan 2004 – Jan 2005 · 1 yr · Boston, MA

• Optical tomography reconstruction:
• Redesigned the existing recursive reconstruction algorithm, increasing its convergence rate and reducing computational requirements.
• Fluorescence Molecular Tomography-Ultrasound prototype:
• Designed and built new hybrid prototype, programming both the acquisition and the reconstruction algorithms.

Jan 2002 – Jan 2004 · 2 yrs · Madrid, Spain

• CT project (MS thesis):
• Designed, built and tested a CT system for small animals. Implemented different cone-beam reconstruction algorithms on a shared-memory system using C++ and OpenMP.
• PET project:
• Implemented real-time data-collection and control routines for a small animal PET system using microcontrollers, custom electronics and two coordinated Unix machines.