This morning RC’er Gonçalo made/showed me and Serena “Bulletproof coffee”. Mad respect to Goncalo. The coffee was tasty and made for a great start to the day. I just googled bulletproof coffee and it seems to be a controversial topic on the inter-webs. I plan on giving it a ‘shot’, actually, you’re supposed to have it in cups..

Pairing with Serena

I got to pair with Serena on her project of implementing her own version of K-means. Serena is going through a paper which explains a different way to go about the clustering/settling (if I remember right) from what is currently done in standard k-means approaches. Serena was very thorough in her annotations, highlighting, and reading of the paper, I was inspired by her due diligence.

Serena and I first paired before check-ins for an hour or so. I was giving her tips on making her code more pythonic. And I learned a lot about the NLP work she’s doing to parse Shakespeare plays, clean them, tokenize them, create feature vectors out of the meaningful text. I’ve implemented k-means before and I have done related work with signals/feature extraction for sound & image processing but I’ve never work with human written text in a data science perspective (aka NLP).

Serena told me I was helpful and she was deriving benefit out of our pairing session which made me happy. She said I wasn’t ‘passive aggressive’. So we paired again in the afternoon. This time I think there was more friction but it was, hopefully, constructive.

I noticed that I had a strong urge to make the code we were writing pythonic, maintainable, testable, and to some extent deployable. This annoyed Serena because her primary concern was to first get it working and that is indeed the right goal for the project at its current stage. I got carried away. I’m wondering if the reason I was trying to make the code ‘production ready’ is because I’m scarred from my last job where I had to write code that was going to be used by a client who will most definitely require changes because they didn’t know what they want. Or maybe I trusted Serena’s expertise in identifying/defining the steps needed to finish the task successfully and thus I had no worries about whether or not it’ll work… Or maybe I wasn’t able to sympathize with the project getting to a ‘working’ state because it wasn’t my project and thus I focused what I found appealing (which maybe means I’m selfish (I don’t want to be a selfish, but we all got flaws that we should identify and fix)… hmm.. I don’t know..

But this interaction raises another question in my mind which is, “how pair-able am I?”. I never thought much about the people skills and techniques needed to be a good and effective pair programmer.. Something I need to look into.. I wonder if it’s possible to measure/match peoples “pair-ability” in the same way sites like eHarmony pair people for relationships…

Pairing with me

The only thing I did independently today is continue my study of ML & Functional Programming. I learned some new things, struggled with some syntax errors because I’m still learning to grok ML stack traces. I’m going to spend this weekend writing a blog post about what I’ve learned so far in ML & Functional Programming.

Talks & Sessions

I made it to the design patterns session and volunteered to study and then explain a pattern in 2 weeks. This will be fun.

I attended Neha’s talk @ Tumblr on the paper “The Scalable Commutativity Rule” written by one of her follow lab mates. There were a TON of brilliant people at this talk. The networking scene was incredible. Everyone was way smarter than I am and the talk was gold!