Building an Open Source Data Science Company (EuroSciPy 2013)

Transcript

1. www.morconsulting.c Building an open source Data Science company EuroSciPy 2013 Ian Ozsvald @IanOzsvald MorConsulting.com
2. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Goals • Excuses for O/S in business (to make a better future) • “How do I start a Consulting company?” • “How do I start a Product company?” • Does your employer make open source? • Who contributes to open source?
3. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 About Ian Ozsvald • HPC/NLP/ML in London for 15 years • MorConsulting.com (9 years O/S only) • Teach PyCon, EuroSciPy, EuroPython • ShowMeDo.com • IanOzsvald.com • StartupChile • Self-published two books (and a 3rd?)
4. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Open Src in business • RedHat (1993) • Enthought (2001) • 10Gen (2007) • Continuum (2012) • Logilab • Elphel, RaspberryPi # talk to Didrik, Peter, Nicholas, Mike
5. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Open Source != Magic Unicorn https://secure.flickr.com/photos/origamiancy/6137631442/ CC-BY
6. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Marketplace for ideas • NoSQL vs SQL • Scikit-learn vs wise.io • psyco/shedskin->PyPy/Numba • Multiprocessing/Disco/PiCloud
7. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Benefits to companies • Visibility, publicity, trust • Easier on-boarding for new users • Improves code quality (tests, docs) • It benefits the community ← big deal! • Easier hiring • Helps with reproducible science?
8. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Benefits to individuals • Demonstration of skills, interests, communication quality • You gain experience (coding/teaching) • 3 billion people coming online...how do you stand out? • A chance to stay ahead during rapid evolution
9. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Licensing • LingPipe have a range of licenses • Please add a license (MIT...) • No license==confusion
10. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Consulting lessons • Do something people want (be specific) • Why do they want a short-term person? • Find 5 paying customers (sanity check) • Bigger customers for longer are better • TALK ABOUT WHAT YOU DO • Partner with good recruiters
11. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Product building lessons • Build something people need • Find 5 paying customers • TALK ABOUT IT • Maximise learning – why would they pay? • Don't code (much) • Start by yourself (take risks early)
12. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 (my) failure modes • Poor market knowledge • Wrong evidence for market (only 1 real customer) • Market too cash-poor (ProCasts) • Projections too optimistic (ProCasts) ->work in a domain you know! • (forgetting to make re-usable IP)
13. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Current experiment: annotate.io • Brand Disambiguation for Tweets (“apple”) • MVP was a 5 page PDF • Super niche->blogging & presenting • Open source MIT (github.com/ianozsvald): • social_media_brand_disambiguator • Goal: Visibility and data licensing
14. Ian@MorConsulting.com @IanOzsvald EuroSciPy2013 Thank You • Contribute and Experiment! • Ian@IanOzsvald.com • @IanOzsvald • MorConsulting.com • Annotate.io • GitHub/IanOzsvald