About
I sit on the investment committee for our Enterprise AI investments, together with our Managing Partner Lu Zhang. A big part of my job is understanding the brave new world of the foundation model, which has raised the bar for all software. Fusion Fund just announced our $190M Fund IV! As always, we seek visionary founders solving big problems with foundational technology. Our typical investment is $1M - $4M in a priced seed round, and we also invest in pre-seed and Series A. Everyone on the investment team at Fusion has technical background and deep operational experience, and most of us have startup experience. Founders can rely on us to have a clue!
Experience
Investment Partner
Fusion Fund · Palo Alto, California, United States
Present
I sit on the investment committee for Fusion Fund's enterprise AI investments, together with our Managing Partner Lu Zhang. We recently announced our $190M Fund IV. Fusion Fund is a seed-stage venture fund based in Palo Alto. We're a technical bunch, and like to invest early. Typical check size is $1M - $4M, often as lead investor; we also write smaller checks for pre-seed rounds. As Fusion's resident data nerd, I'm always looking for startups focused on machine learning, business intelligence or data infrastructure. As AI innovation continues at a dizzying pace, I spend a lot of time figuring out what technology will actually be useful. Our enterprise AI portfolio includes Vectara, Otter.ai, You.com and Huma.ai. Everyone on the team has technical and operational background, including our Managing Partner, who holds an MS from Stanford and founded a medical device startup. Founders can rely on us to have a clue!
Principal, Jll Spark Global Venture Fund
JLL · San Francisco Bay Area
The JLL Spark Global Venture Fund launched in June 2018, combining the investment upside of early stage proptech with the massive distribution of a Fortune 500 company. - Led JLL Spark's efforts for our Series A investment in OpenSpace, which later raised a Series D at a $902M post-money valuation. Developed investment thesis and negotiated terms with CEO. And got quoted in Forbes! https://bit.ly/2YWITSc - Led JLL Spark's efforts for the SwiftConnect Series Seed, which we co-led with Navitas in October 2021. I handled every aspect of the transaction, including sourcing (thanks Union Labs!), due dili, writing the deal memo, negotiating the terms and drafting the term sheet. Here's the press release: https://bwnews.pr/3MifiMM - Led JLL Spark's efforts for our Series B1 investment in Eden (eden.io). We were the lead investor in this $29M financing, our largest investment at the time. Developed investment thesis and negotiated terms with CEO. Herded multiple cats (due dili, legal, JLL inte…
Vice President, Jll Spark
JLL · San Francisco Bay Area
JLL Spark started in June 2017, when JLL CEO Christian Ulbrich hired Mihir Shah and Yishai Lerner to lead a new group focused on commercial real estate technology ("proptech"). One month later, Mihir and Yishai hired me as "Employee #1". Learned the commercial real estate business and identified areas of potential tech disruption. Worked with JLL Spark leadership to determine the best operating model for this new division. Created a machine-learned model to predict residential rents in major US cities.
Vice President of Data Science
BigML, Inc · San Francisco Bay Area
The Remington shaver approach: I liked the product so much, I joined the company! Represented BigML at industry events. Gave live demos of the BigML tool using real datasets. Explained our approach to both technical and nontechnical audiences. Venues included CERN, ML Prague, SVForum, DataBeat, DataWeek, Zipfian Academy, Big Data Spain, Papis.io, the LA Machine Learning Meetup and the SF Machine Learning Meetup. Wrote blog posts explaining anomaly detection, predictive modeling, decision tree visualization, the concepts of training versus scoring, and bootstrap aggregating. Sourced new customers and helped existing customers understand how to use BigML. Conceived new features to improve the BigML product. Features that launched included Tableau model export, an interactive confusion matrix and enhancements to the SunBurst visualization. Wrote Python scripts to train and evaluate decision tree ensembles using the BigML API. Varied model parameters to improve accuracy, and validated resu…
Data Investor
My humble self · San Francisco Bay Area
Early stage investor and advisor focused on data companies. Advised VC investors on Platfora, Pepperdata, SiSense, Interana and BigML. - Lead angel investor in BigML and member of the Strategic Advisory Board. - Seed investor in Mob.ly, acquired by Groupon. - Seed investor in Alation, which raised a $110M Series D at a valuation of more than $1B in June 2021. - Series A investor in Cloudera, Pepperdata, Interana. - Series B investor in SiSense, which raised a $100M Series F at a $1B valuation in January 2020. - Series B investor in Platfora, acquired by Workday. - Member of the Scientific Advisory Board for Hi Fidelity Genetics.
Director of Data Science
Groupon · San Francisco Bay Area
Built an elite data team of 9 people (including 3 quantitative Ph.Ds), each with a rare combination of analytic brainpower and hands-on data systems skill. Through attentive people management, retained all 9 despite stock volatility in 2012. Finished Groupon with a perfect retention record. Team created the first machine-learned deal relevance model for the Groupon mobile app. Used hypothesis testing of means to analyze iPhone and Android A/B tests, giving the Mobile team confidence to launch new apps with radical improvements to the UI. More than 3 years experience writing intricate SQL queries on Hadoop, Teradata, Vertica and (shudder) MySQL. Used MySQL production replica to create the first flattened tables for business use. Devised hacky solutions to deliver data to the business, buying time for the creation of a proper data warehouse. Delivered real-time Tableau reports used daily by 100+ business users. Deciphered Groupon's web database schema and translated into easy-to-understa…
Principal Program Manager, Bing
Used Microsoft's analog of Hadoop to analyze extremely large data sets. Gave lectures on the SQL interface, MapReduce-style computation framework and underlying distributed filesystem. http://bit.ly/NGXxoQ Co-authored short paper "Detecting Uninteresting Content in Text Streams", accepted at the SIGIR 2010 crowdsourcing workshop. Used human labels from Amazon Mechanical Turk to train a simple classifier that identifies "uninteresting" tweets with more than 80% accuracy. http://bit.ly/MyUEo7 Worked with engineers to load Twitter firehose into production MapReduce cluster. Worked with researchers to load Twitter firehose into experimental MapReduce cluster. Wrote scripts using SCOPE, a SQL interface analogous to Hive for Hadoop, to answer important business questions. http://bit.ly/lTbs9 Performed search ranking experiments and analyzed relevance boosts from improved training data. http://bit.ly/8BCC14 Used "click labels" to train ranking functions in Bing international markets. Experime…
Director of Analytics
Conceived, designed and launched collection of URL "clickstream" data via Yahoo Toolbar. Wrote product spec and worked with research, engineering, data systems, policy, legal and strategy teams to implement. As a result of this project, Yahoo has logged tens of billions of clickstreams in its Hadoop infrastructure, with billions more collected each month. Yahoo still uses this data today to understand user behavior and optimize its advertising business. Collaborated with researchers to identify promising clickstream features for the search ranking algorithm. Multiple features launched in the algorithm, improving relevance. Worked with web performance expert Steve Souders to include page load time tracking in the Toolbar clickstream project. As a result, billions of page load times are now recorded in Yahoo's Hadoop infrastructure, representing one of the largest repositories of actual page load times measured from the client side. (Page load time was also useful as a feature in search…
Principal Technical Yahoo!
Managed a team of seven analysts. Responsible for A/B testing on Y! Shopping. Implemented test process, tested new product features and evaluated impact on revenue and user experience. Devised and evangelized a method for improving Yahoo's search and sponsored search relevance. Successfully lobbied for resources to implement this "20% project". Managed business intelligence for Y! Marketplace properties, including bucket testing for Yahoo! Shopping, Autos, Maps and Local. Managed the MicroStrategy engineering team for Y! Shopping, Autos, Maps, Local, Travel, Personals and Real Estate.
Manager, Search Monetization
Managed A/B testing for Yahoo's main search page. Sponsored search ads on this page generated revenue of more than $1 billion per year. Used A/B testing to prove the value of a new sponsored search matching algorithm and overcome internal resistance to the change. The new algorithm increased sponsored search revenue by $80M per year.
-
David Gerster Structured Finance Services
Created financial models for securitizations. Clients included KPMG, Wells Fargo and DaimlerChrysler. DaimlerChrysler engagement involved securitizations worth more than $20 billion. Please visit http://www.goCMO.com for more information. Keywords: Securitization, structured finance, mortgage-backed securities, MBS, asset-backed securities, ABS, commercial mortgage-backed securities, CMBS.
Founder and Ceo
DeltaClick, Inc.
Raised $2M for Internet startup focused on contextual advertising. More info at http://www.gocmo.com/deltaclick. Responsible for all product design and feature decisions for both client and server. VP Engineering and Chief Software Architect reported directly to me. Product was installed on more than 1,500,000 machines and received high ratings from users.
Education
University of California, Berkeley, Haas School of Business
MBA
Harvard University
BA with Honors, General Studies
Nichols School
High School
Find leads like David
Origami finds your ideal customers across 15+ data sources with a single prompt.
Try Origami Free