From Research to Industry, in Ten Not-So-Easy Steps(Bridging the Gap Between Academia and Industry)
Radim Řehůřek, Ph.D.rare-technologies.com
2005
2007
2010
2011
Respect for open source?
http://wstein.org/talks/2016-06-sage-bp/bp.pdf
RaRe Technologies Ltd.
2016 – Academic partnerships
#1 Every company is a shit show
#2 Nobody cares about experts showing off, or ivory towers
“The purpose of computation is insight, not numbers.”
Richard Hamming● clear processes: scoping, error analysis● always eye toward business impact
○ vs. technologist mindset, tinkering for fun
Pragmatic applications of academic results worth gold!
http://rare-technologies.com/performance-shootout-of-nearest-neighbours-querying/
#3 Pricing● beware of hidden costs of consulting!● expect to bill ~50% hours: not 1:1 salary
○ the rest: keeping up with tech, research○ prospecting, accounting, proposals, marketing ○ internal trainings, code reviews○ +open source “free support”...
● charge for business analysis!○ that bit to understand client’s business, goals & data○ before any implementation and before SoW
Cambrian explosion
“Do one thing and do it well.”Doug McIlroy
"Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can."
Zawinski’s law of software development
#4 Master your tools
Business value >> tech stack
http://www.slideshare.net/ryanorban/how-to-become-a-data-scientist
Lots of existing resources
#5 Prepare for difficult, poorly scoped “challenges”
Good● get experience w/ various
verticals, industries● learn the process: how to
scope, price, analyze, deliver
● develop ancillary skills: accounting, taxes, IP law...
Bad● uncertainty not for
everyone● transfer of domain
expertise difficult● expectations to master
all new tech instantly, constantly
#6 Take progress easy
● paper into production = massive investment!
#7 Expect non-glamorous work
1. problem discovery, scoping, APIs2. data collection, cleaning3. evaluation, reporting, visualizations4. coding, integration, automationvs. what novices expect: models, math, algorithmsvs. suits & armchair AI philosophers
#8 Avoid wasting time on “unrealistic” clients
Bad● enterprise “demonstrate innovation
to execs” projects... unsatisfying● idea stage start-ups “golden
goose… on $5k budget!”... naive● veiled attempts at hiring... a
flattering distraction
Good● smaller/medium companies with
established market fit● reporting to CTO / CIO / project lead● ideally longer term collaboration:
sense of ownership
How?● Price yourself out of “I have an idea” range● Manage expectations clearly, terminate dead ends quickly
#9 Be wary of working for equity
● too risky○ do you really want to be an early investor?
● core UVP tech designed and developed by an external consultant... tricky○ esp. remote consulting
● leverage & impact mismatch
Navigating the hype● It’s your responsibility to
manage expectations and advise on tech stack
● Diplomacy needed: “big data” & co
● Cultural differences
Rare Brain Trust
#10 Scaling up consulting sucks
Good● people help scaling up● take on larger projects,
with larger teams
Bad● people don’t help scaling
up● gold rush, “instant
billionaires”
...Teach yourself “data science” in ten years
Bonus tip #11: take time to appreciate your luck
Radim ŘehůřekRaRe Technologies Ltd.
http://rare-technologies.com(previously radimrehurek.com)
@radimrehurek
Blog (mostly tech+machine learning, some travel):http://radimrehurek.com/blog/