What I did in 2022
I did a LOT in 2022. Way too much, especially in December when I was still a few months into a new job, running a conference, wrapping up a compsci class, and trying to plan a family trip to Argentina. Note, none of this does includes my other main hobby - actively parenting two small children.
As soon as Normconf was over, I got so sick I was in bed for two days straight. It was a cold my kids brought home from school and daycare, but also my body’s way of telling me I need to slow down or I’m going to die.
So, this year, I am actively turning down any asks that do not relate directly to working on Viberary and slowly regaining my sanity.
- Took some time over the holidays and migrated this blog to Hugo. Blog migrations, regardless of what you’re migrating from or to, are always an ENORMOUS pain where you lose content and structure, and I hope I don’t have to do it for another ten years.
- Wrote about three tools I think everyone in data should know: git, cli, and SQL. I have been working up to a theory for the past couple years, and my theory is that good data work is just good engineering practices, so learn to be a good engineer as a data person, and you can do anything you want in the data space.
- Worked some on boringml, which is just a bunch of notes on the non-sexy stuff in recsys/ML. It’s still in pretty rough shape, but hopefully I’ll have some more time to flesh it out.
- Wasted a lot of time trying to understand sbt and optimize it
- Wrote up unit testing for Scala Spark based on work I did at Tumblr in standardizing some of our machine learning project code
- Lots of work at work at Tumblr on personalized recommendations, content classification, Airflow, ranking, and filtering kept me pretty busy
- Co-organized Tumblr’s Feeds and Recommendations first-ever meetup in NYC
- Wrote about how multi-armed bandits work
- Along with my team, completed my first paper submission to Recsys 2022. It did not get accepted.
- Read and endorsed “Essential Math for Data Science”
- Wrapped up my job at Tumblr
- Started working on Nisaba, a Telegram bot that would write out links to a web site that I could search through, like Pinboard. My main method of consuming and reading links is to send them to “saved messages” in Telegram, which is ok at searching, but I’d really love to also be able to tag and filter better. Because I got too busy I never finished this project, but would love to come back to it because I’m still sending links to Saved Messages and reading them.
- Started a new job at Duo, working on building out machine learning platform infra
- Went to Recsys 2022 virtually and took a ton of notes that I still have to write up. I have developed a system in Google docs to do this.
- Wrote about recommendation systems and how they work in social feeds in response to the growing chaos at Twitter
- Recorded a podcast with Peter Wang of Anaconda about Human in the Loop machine learning. I’m not sure how, but Peter knows everything about everything and it’s an extreme pleasure to talk to him.
- Started taking a college-credit class on OOP and Algos in Java. As a side note, community college in the US is an extremely underrated way to learn stuff and still be accountable because you pay for real college credit
- Went to a talk on HNSW at Pinecone and took notes
- Gave a talk at Meta about contributing to PyTorch (thanks to Mark for the invite!)
- Took Grant Ingersoll’s excellent Search Fundamentals Course to better understand search, which I’d always worked parallel to but never really introspected as much as I should
- Got a new project idea: semantic book recommendations, and started working on Viberary
- Stable Diffusion hit and I started learning more about its internals
- Did some reflection on the nature of remote work and rituals and how we need to be deliberate about building them
- Re-read Effective Python
- Started learning about Mastodon’s infrastructure
- Wrote about how I learn ML
This was a Bad month and I have blocked it out of my memory.
- ChatGPT hit like a hurricane and I spent some time understanding where it came from and what it was . I’d previously written about the business context of OpenAI so this was particularly interesting
- Finished that class on OOP and Algos in Java, complete with a Java static site generator
- Finished reading Learning Java in tandem with my class
- Read and edited the English translation of antirez’s Wohpe
- Played around with Pyscript
- Wrote a post I’m pretty proud of, about programming in the cloud, related to digging into what I’ll need to do for data storage for Viberary
- Normconf Happened
- Gave the keynote talk of Normconf
- Did a retro of it