Profile

Cover photo
Felipe Hoffa
Works at Google
3,055 followers|901,856 views
AboutPosts

Stream

Felipe Hoffa

Shared publicly  - 
 
Open data unleashed: The NYC taxi dataset hackathon

The NYC Taxi and Limo Commission released last year 174 million taxi trips from 2013 through a Freedom of Information Law request. In this video, they unveil the 2014 data on a historical date at the NYU Rudin Center for Transportation:

For the first time in the TLC history, data scientists, civic hackers, graphic designers, analysts, government policymakers, statistics hobbyists, and taxi drivers gathered to discover solutions to the taxi gap, which is estimated to miss out on about 11,000 rides daily.

Discuss: https://www.reddit.com/r/bigquery/comments/3d4ljg/video_open_data_unleashed_the_nyc_taxi_dataset/

NYU recap: http://wagner.nyu.edu/rudincenter/2015/04/taxi-shift-change-hackathon-event-recap/

2013 data and queries: https://www.reddit.com/r/bigquery/comments/28ialf/173_million_2013_nyc_taxi_rides_shared_on_bigquery
7
5
Alex Osterloh's profile photoMandy Waite's profile photo
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
BigQuery UI secrets #2: You can auto-complete (press TAB)
18
4
Felipe Hoffa's profile photoRoberto Landrau's profile photoGerry Alvarez's profile photoRatna Koppaka's profile photo
2 comments
 
it's been there for a while :)
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
 
This is the most fun we've had in the office in a while. We've even made some of those 'Inceptionistic' art pieces into giant posters. Beyond the eye candy, there is actually something deeply interesting in this line of work: neural networks have a bad reputation for being strange black boxes that that are opaque to inspection. I have never understood those charges: any other model (GMM, SVM, Random Forests) of any sufficient complexity for a real task is completely opaque for very fundamental reasons: their non-linear structure makes it hard to project back the function they represent into their input space and make sense of it. Not so with backprop, as this blog post shows eloquently: you can query the model and ask what it believes it is seeing or 'wants' to see simply by following gradients. This 'guided hallucination' technique is very powerful and the gorgeous visualizations it generates are very evocative of what's really going on in the network.
35 comments on original post
2
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
How it feels when people don't want to try BigQuery.

(credits at http://www.reddit.com/r/bigquery/comments/38d8ty/humor_how_it_feels_when_people_dont_want_to_try/)
29
4
Jason Morrison's profile photoJean Joseph's profile photo
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
Picture this: Briefly after meeting +Jaclyn Kollar, she mentioned she had her genome sequenced by 23andMe (so genotyping). I told her I knew how to use Google BigQuery to explore it - and so we did!

The future is now.

Google Genomics has a full cookbook on how to do this: http://googlegenomics.readthedocs.org/en/latest/
24
5
Charlie Wong's profile photoPeter Mancini's profile photo
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
 
Cloud Dataflow in beta: already better than lambda in production?  Time to stop writing your code twice, in two different systems, to get both batch and streaming analytics.

(http://lambda-architecture.net/ if you're not familiar with that term.)
3 comments on original post
1
1
Maxim Ostapenko's profile photo
Add a comment...
Have them in circles
3,055 people
Pamela Rocha's profile photo
Ivo Limmen's profile photo
roberto miranda's profile photo
Higinio Juan Castro Rodriguez's profile photo
Marco Pöhler (mpolo)'s profile photo
Maria Francisca Rodríguez's profile photo
Eduardo Ibáñez Quezada's profile photo
Manuel Bahamondez Honores's profile photo
Felipe Baytelman's profile photo

Felipe Hoffa

Shared publicly  - 
 
Visualizing 8 years of Reddit comments: A short video!

https://www.youtube.com/watch?v=l8MLIfU21pk
7
1
Daniela Ugarte Cisternas's profile photoJason Baumgartner's profile photoFelipe Hoffa's profile photo
2 comments
 
Great work!  This really illustrates the power of BigQuery when applied to social media.  There is just so much that can be done with this data.
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
 
We recently switched from Godaddy dedicated servers to Google Cloud Compute VM servers and due to that, our system was running nice and cool during this record usage. Here's the CPU track showing a maximum of 27.5% utilization at the time of the record. Our (former) Godaddy server would have collapsed under this load. Godaddy doesn't offer a server powerful enough to handle our requirements. Google Cloud Compute offers plenty of power for about the same cost as the maxed out Godaddy system offered.
2 comments on original post
4
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
An in depth article on how to map the world news thanks to the new CartoDB+GDELT partnership. It even includes my BigQuery example :)

http://blog.cartodb.com/gdelt-api/
In the second addition of our series on clever mapping with GDELT data (see this post from last week), we wanted to take on some more advanced mapping t...
6
6
Deborah Lee's profile photoAlex Gordon's profile photo
Add a comment...

Felipe Hoffa

Shared publicly  - 
3
2
Felipe Hoffa's profile photoAlex Gordon's profile photo
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
An awesome hackathon this June in Amsterdam! Join if you can (or tell your friends that might come)
 

BigQuery + OpenData hackathon, June 2-4

My buddies at Appsterdam are helping organize the Amsterdam Smart Cities hackathon June 2-4. It's cool for all the reasons hackathons are usually cool: cool location INSIDE the Amsterdam Arena, European-wide exposure, several cool IoT products in the mix (the lovely people at Glimworm iBeacon, and a neat early-startup called Trakkies)

Why should you be excited, other than just being a fun event?

Because I've worked with Tom (Appsterdam) and Jasper (Amsterdam Economic Board) to arrange that all Open Datasets are unified and available in BigQuery. That means you can do your mashups, exploration and analysis over all datasets in the same place. I'm pretty excited about this, there are some Open Datasets in BQ but this is the first time I've seen all data for a Hackathon in one unified place. This includes static data, but also realtime data from OV, Trakkies, energy sensors, more.

You'd start by exploring the data in the BQ web console (just write SQL), find some cool correlations and queries, then throw it into a visualization layer for cool charts, or write a website or mobile app with the many BQ libraries (Node.js, Java/Android, Go, .NET even!)

Signup at the Appsterdam meetup page!
Hack the City and the Amsterdam Arena! Join us for the first Smart City Hackathon at the famous Amsterdam Arena. This is a unique opportunity for programmers, designers & business developers to solve real issues of smart energy, smart mobility and event experience. We are co-organizing this hackathon with Appril. The event partners KPN, Huawei, FAN, Accenture, the City of Amsterdam and the Amsterdam ArenA are offering full access to open & live...
View original post
2
Add a comment...

Felipe Hoffa

Shared publicly  - 
 
How to create a heatmap of worldwide coverage of news for the last 24 hours (in less than 1 minute!)

GDELT+BigQuery+CartoDB
13
23
Nicholas Tan's profile photoNash Maruvada's profile photo
Add a comment...
People
Have them in circles
3,055 people
Pamela Rocha's profile photo
Ivo Limmen's profile photo
roberto miranda's profile photo
Higinio Juan Castro Rodriguez's profile photo
Marco Pöhler (mpolo)'s profile photo
Maria Francisca Rodríguez's profile photo
Eduardo Ibáñez Quezada's profile photo
Manuel Bahamondez Honores's profile photo
Felipe Baytelman's profile photo
Work
Occupation
Developer Advocate
Employment
  • Google
    Developer Advocate, 2013 - present
  • Google
    Software engineer, 2011 - present
Story
Tagline
From Chile. Working at Google. Living in San Francisco.
Bragging rights
"I find a duck's opinion of me is very much influenced by whether or not I have bread." --MH