AJM Capital sources off-market apartment communities in emerging markets backed by data science. We partner with experienced operators to reposition apartments with value-add strategies and co-invest with accredited investors.

Platoon x  Music

• Machine learning and data science consulting for high-growth startups and enterprises.

• Lead data science initiatives for the Incubation business unit (65 people).

Built petabyte-scale supervised machine learning systems to predict event ticket sales for 2M+ events daily, improving model accuracy by a factor of 3.5x from a naive baseline model (MAE). Collaborated with data engineers to deploy Eventbrite's first in-product ML models.

• Exited in February 2018 through acquihire and IP sale to Eventbrite.• Built end-to-end supervised machine learning system to predict concert ticket sales for the live music industry with 9% error margin (MAPE), topping 21% industry average. Data sources: music streaming services, social media, geo-demographics (census), Wikipedia. • Applied research under Professor Matthew Turk: paragraph vector embeddings (doc2vec), generative probabilistic topic models (Latent Dirichlet Allocation), hierarchical cluster analysis (Ward’s method), non-convex clustering, nonlinear dimensionality reduction (t-SNE). • Contracted data partnership with major music industry company under NDA.• Designed and developed full-stack web app dashboard prototype.• Tools: Python, pandas, numpy, scikit-learn, seaborn, TensorFlow, Keras, gensim, IPython/Jupyter, AWS, SQL, Flask.

• Helped develop C++ module to integrate sFlow network packets (pcap) into standard NetFlow protocol for Cisco-funded startup in Seville, Spain. Tools: C++

• Worked with the Core Applications team to develop an internal tool visualizing SCXML files.• Developed uSCXML test-runners with GUnit, Google's open-source testing framework.• Tools: JavaScript, jQuery, HTML/CSS, C++