The Parallel for Company Building (Notes from Dec. 2 - Dec 8, 2019)

Keeping this short because I'm working at putting the idea together. Love the parallel for a gym atmosphere and company building.I bet it's already forming in your head. Imagine the big class at a big box gym, the personal ones, classes, aggressive combat, cycling, different personalities in weight lifting, cardio spending and all the pairs of trainer + clients you can imagine. It's in the works!I do hope you enjoy the deep dive and notes I took in hearing Buck Woody's AI podcast. You should also review Louis's app for Total Brain. Wrap that up with Dustin Dolginow's review on how to utilize the power of the internet to own the VC and investing interest.

Week of December 2, 2019

  • Louis Gagnon, CEO, Cofounder of TotalBrain (Work and Life, WhartonXM)
BRAND •    Medium size • No tagline • v2 copy
  • Talking about each person having different brains, therefore different treatments
  • Making sure that the brain is merely a part of you, not all of you - racing In these years, it's a large amount of stress - good or bad
  • Buck Woody (@buckwoodymsft), Applied D/S at Microsoft (Data Skeptic 12/3/19)
acastro_180507_1777_microsoft_0001.5
  • Lots of intricacies in bonsai trees
  • Last thing you get to learn is the watering can - easily under/over-water trees
  • Scissors as the primary tool
  • ML/AI/BI and advanced data analysis - rigor of right spread, amount, representation, statistics things are the watering can If base data is wrong, the model is pointless
  • He works in the industries teaching classes on MS platform for SQL Server (which contains Spark and many other things), data science
  • Simply looking at the data - where did I get this? Say, financial projections: financial data (how do you know it's here)
  • Predictions as pedantic/boring: preventative/maintenance predictions - wanted to ensure units wouldn't fail
  • Half the time, predictions worked well, half the time, worked awful
  • How did the data report from the machines? Had to go to manual of machine (there was older, newer) to see data
  • Anomaly was that the older machine was reporting every 60 minutes, not minutes (which the newer was doing)
  • Works with many users - fraud and anomaly detection Use case example of gaming company with cheating and making sure the data was good
  • Regionalized languages - programming as how you think of your solutions Big things to do: Kubernetes and Containers - be very familiar with environments to make sure infrastructure is done well
  • Looking through data science process - who wears what hats, data engineers and DBA having overlapping roles
  • Many he comes across that don't know that database guys can do much of what they're looking for
  • Often the requirement, if given to data scientists late, will be multiple projects
  • Containers run-time - (docker) vs Docker
  • Text file (yaml) with Python 3.5, MySQL and code as file - compose into an image (gathered up version of those runtimes)
  • Tell Docker to run it - container - description to image to container - not representing memory and disc, just using that on station
  • Docker smart enough to recognize that it will run similar versions
  • Kubernetes - KAS Another yaml file and engine, on a node (physical or virtual) with docker runtime, couple of services (networking, part of cluster) with

Master node that makes sure everything is happening - wrangles everything for you in a persistent volume for the pod since storage was an issue

  • Thinking of SQL as declarative language - select * from mytable isn't what we do
  • Containers are declarative language for computers, essentially
  • Kubernetes is the platform or network for a full declarative network
  • Business intelligence in 90s - specialized people as parts of it with only some people knowing how to use, prechew for users and it took months
  • 5-6 years ago, this still remained - data scientist would spend 99% of time in economic data or weather data or whatever model, version or experiment Walk out with tablets, thus save the data - maybe another she was working with, maybe not
  • Data engineer is most sought after job title - "everything but the algorithm" at Microsoft (LinkedIn)
  • Link - aka.ms\tdsp, defines out team structure for data science team with guides - devops, mlops, aiops, mlops
  • People are used to BI projects - one cube, answer lots of questions but with a data scientist, if you question "this and this and this.." - separate 2 or 3 different data sets, can't answer clustering question with regression algo, how many of these vs which things do they belong to
  • For large orgs: Do you know if you have a DB team? All data in its forms.
  • Showed someone SSIS done after a minute after they started pulling up R and his algorithms - "wizardry"
  • Used to work at NASA, talked of a friend scientist who landed a round camera on the moon ahead - had to turn it away from sun because it'd melt film
  • Some of cast of Star Trek would show up all the time at NASA, large glass rooms (lab coat, tie, white shirt)
  • James Doohan, Scotty was gonna show up. Scientists would go to break room and watch Star Trek in 60s (debate whether or not stuff was possible) Mentions automatic door as someone being off camera in the 60s, possible or not on physics
  • Taking notes and turns camera back on - "Fascinating" from Leonard Nimoy on the outside of glass
  • He wants to make people know that they have people that can fold into data science team Cultural that DBAs think they'll need to be data scientists and data scientists that are territorial (don't want people messing around)
  • Young: computational basics, logics, data processing
  • High level math - stats/linear algebra
  • Domain expert: particular vertical like healthcare, finance or patterns available for a width of an application of a tech
  • Learning to learn: how to pick up and put down knowledge - pace of learning something (can't be an expert in the timeframe) Pick language you like and then figure out how you're learning it - then, do it again for others
  • Hours of studying that can be pre-chewed - lack of focused time, spend too much time on all of it Confs where people get away to focus on a topic (until they get on their phone and blow it)
  • Where's AI going? He says - going away. "Nobody says they have computers at the company anymore"
  • "I" or "e" in front of company and get funding anymore, same with cloud - just ubiquitous, computing/drive
  • Predictive/prebuilt AI now - text analysis, image processing, predictions
  • Need to know how to trust it or trusting it too much - aka.ms\ai-ethics
  • Flash fill, for instance, in Excel - Microsoft Research done in PROSE AI in the cell, disappears into product
  • Ex: PowerPoint presentation coach with mic on and it will critique you
  • Dustin Dolginow (@dolginow), GP at Maiden Lane (20min VC 2/10/16)
ttcp_web_portfolio-update-logos-small_230x150_maidenlane
  • Online venture fund using AngelList as its o/s, capital partner to best angels in the world, investments in Getable, PipeDrive, Beepi Also venture partner with Accomplice, was previous operator at Social Swipe allowing merchants to gain value from txn data
  • Went to college in East Coast, Wall Street at Lehman Brothers during crash and decided to do a product idea in payments for 1.5 years Product dev, front-end and shifting from finance world - introduced to partner Jeff (running Atlas Ventures, renamed to Accomplice)
  • Started taking introductions to companies as Nidhi, Naval and AngelList would be giving them - since 2010 and normal user
  • Atlas lead the AngelList series A and every round since - 2012 moved to SF and make VC legal - 2013 for syndicates start Lead a syndicates fund only in 2013 with Jeff - learn by doing and figure out what it meant - $25mln named for Maiden Lane in SF Irony was AngelList HQ was on it, one of bailout funds for Goldman Sachs real estate was Maiden Lane
  • Figured Syndicates could be impactful for institutional investors, also
  • Moved to SF in 2014 to close the fund, April started investing in the fund
  • Native app on AngelList - (like saying Uber is Apple because on Apple) - put in their docs 50% off-A/L, 50% on but realized it was moving quickly
  • AngelList as unbundling the activities of VC - funds are containers for capital / infrastructure AngelList has more flexibility but it's a small container - box - they're doing product first with data that it creates
  • Working with set of angel investors that take their money and invest on their behalf - share carried interest within GP's (30% carry, no mgmt fee)
  • 15-20% of carried interest goes to syndicate leads - driving brand, operating within company, adding value, interacting
  • His goal is to shepherd to create resources for community (most syndicate leads have other jobs) - live work loft, for instance
  • Community and flexibility is a big part of it - consensus-based decisions, non-consensus (conviction for solo), LPs as direct investments for bigger broader
  • Entrepreneurs as understanding users' needs - great community done by Ryan Hoover at Product Hunt
  • Most overhyped - prescription delivery, underhyped - Canada as country, developer tax credits
  • Goal for Maiden Lane - kickass set of syndicate leads that get called upon by lane
  • Last impression from a book: Development is Freedom by Amartya Sen
  • AtVenu - as most recent investment