Jun 2019 – Mar 2021 · 1 yr 9 mos · Los Gatos, CA

• Led the Machine Learning Infrastructure team and our open-source product, Metaflow.

Aug 2017 – Jun 2019 · 1 yr 10 mos · Los Gatos, CA

• Designed and developed the first versions of Netflix's new human-centric machine learning infrastructure, Metaflow, which is now open-source at https://metaflow.org

Jun 2016 – Aug 2017 · 1 yr 2 mos · San Francisco

• Led various data and machine learning efforts at AdRoll.

Aug 2015 – Jun 2016 · 10 mos · San Francisco

• Grew Prospecting to drive tens of millions of dollars of revenue working closely with business stakeholders.
• Open-sourced traildb.io, a core component powering Prospecting and other data-intensive systems at AdRoll.

Jun 2013 – Jul 2015 · 2 yrs 1 mo · San Francisco

• Hired and led the first analytics team at AdRoll. Built a custom in-memory, JIT-compiling data store to power the analytics stack.
• After this, led a team that designed, built, and operated AdRoll Prospecting, a new customer acquisition product based on PB-scale machine learning.

Sep 2011 – Jun 2013 · 1 yr 9 mos · Palo Alto, CA

• Co-founded a startup that built a novel scriptable data platform. You could define your own metrics and ML models in Python, using a slick in-browser IDE, while our backend took care of heavy lifting on AWS.
• I was the CEO. Bitdeli was acquired by AdRoll in June 2013.

Jun 2010 – Jul 2011 · 1 yr 1 mo · Palo Alto, CA

• Managed and grew a research team focusing on big data infrastructure, analytics, and machine learning.
• Our skunkworks developed and operated the largest, petabyte-scale data repository at Nokia using Disco. We solved data challenges from basic KPIs to predictive models and visualizations.

Sep 2007 – Jun 2010 · 2 yrs 9 mos · Palo Alto, CA

• Authored Disco (http://discoproject.org), an open-source implementation of MapReduce written in Erlang and Python. I helped to grow a small community around the project that has been active ever since.
• Implemented Ringo (https://github.com/tuulos/ringo) that was one of the first open-source NoSQL databases designed after Dynamo, predating Cassandra and Riak.

Nov 2006 – Dec 2008 · 2 yrs 1 mo · Helsinki, Finland

• Founded a data science (before: business intelligence) consultancy.

Jun 2004 – Sep 2007 · 3 yrs 3 mos · Helsinki, Finland

• Worked with a research group at University of Helsinki. Our research included Bayesian modelling, information theory, probabilistic information retrieval, multidimensional visualizations, and large-scale computation.
• Some of my fun projects include a probabilistic web search engine (https://github.com/tuulos/aino), a large-scale urban game in NYC (http://www.youtube.com/watch?v=3ozUNUTNMT4) and 10+ papers (http://www.hiit.fi/cosco/publications).

Dec 2000 – May 2004 · 3 yrs 5 mos · Helsinki, Finland

• Gurusoft was a startup founded to commercialize self-organizing maps (http://en.wikipedia.org/wiki/Self-organizing_map), a peculiar kind of artificial neural networks.
• We promised to revolutionize intranet search and document management by making it visual and probabilistic. The idea was way ahead of its time.

Jun 1998 – Sep 2000 · 2 yrs 3 mos · Oulu, Finland

• Developed systems to classify lumber, coffee beans, and movies based on computer vision.