2021 has been kind to me. After moving to Seattle (from Singapore) at the start of 2020, I’ve enjoyed my work at Amazon where I help customers discover books and read more, and learned a ton in the process. My wife and I had the chance to visit a few US states as well as enjoy Seattle’s beautiful nature. My immediate and extended family has stayed healthy. This peaceful backdrop allowed me to make progress on my goals, as well as take on unplanned “projects”.

Past review: 2020

2021 goals—hits and misses

Write 48 posts, including six informal mentor and six teardown pieces. Excluding this post, I’ve written 34 posts this year—thus, the overall goal was a miss.

That said, part of the goal was six teardowns and six informal mentor interviews—this was a success. I wrote seven tear downs, mostly on recommendations systems. These took more effort than regular posts given the research and organization required, but I also learned a lot writing them. I also started a new project and got nearly two dozen informal mentor interviews and made just as many new friends.

Give 4 talks. I gave 10 talks/podcasts this year and met the goal. Along the way, I learned that I didn’t enjoy giving talks where I was mostly repeating myself. I spent too much time preparing slides that were largely summaries of what I had previously written about. Will probably scale this down in 2022.

Receive 100 “this is awesome/helpful” feedback. I hit 100 sometime in June or July and stopped counting.

3 - 6 mentees achieving the next level. A handful of mentees and folks that reached out for advice grew in 2021, either by getting promoted and getting bigger scope, or transitioning into their next role/dream company. Thus, I consider this a success.

1 - 3 start-ups invested in/advising. Several founders reached out for advice and feedback, and offered the opportunity to angel invest. (They learned about me via my writing or VC introductions.) I invested in two that I’m excited about and want to see succeed—one in the ML production/marketplace space and one in the cloud infra space. They’ve made tremendous progress in the past couple months and I’m looking forward to more. I also made a handful of angel investments via AngelList as an LP.

Meditate 60 minutes daily. I was consistent till July. Then, I adopted a puppy (more later). Now, she requires all my attention in the mornings. Will want to get back to this once she’s grown up and mellows a bit.

Learn to snowboard. Nope. Still want to do this sometime though.

Learn TypeScript/JavaScript. I tried learning via courses but didn’t get very far. Though I completed 30 days of JavaScript (and did it in TypeScript), it was boring. I tend to learn better by applying along the way. To make it more fun, I started a new project (more later) where I learned how to apply basic JS, linting, and building fun features with JS and React. Thus, I’ll consider this a success.

Eat out twice a month. Yes! And this got easier as Seattle opened up. My tummy is happy.

Read six fiction/pseudo-fiction books. Currently 0/6. My reading choices have been more practical, where I focused on papers and books I can apply immediately. Maybe over Dec?

Visit three US states. 2/3. Visited Los Angeles in late spring, and then Portland in November for some hiking and shopping (upgraded my iPhone 7 to the iPhone 13 mini and I love it).

Other than these planned goals, a few other serendipitous events happened.

Adopted a rescue puppy

Early this year, my wife and I dog sat for the first time. We took care of the most adorable, well-behaved dog that just wanted to fetch non-stop. Over those nine days, we built confidence that we could take care of a dog and thus decided to adopt a rescue puppy.

Latte growing up in 2021

This is Latte. If you’re subscribed to my newsletter, you would have heard me mention and share a few pictures of her. She’s slightly over six months old now and is affectionate, brave, smart, and loves people more than dogs. She’s also very mischievous and requires 100% supervision or she’ll be up to no good such as chewing on my Monstera or cables.

She’s upended my morning routine. Previously, it was meditation, writing, and breakfast. Now, it’s her potty, her breakfast, training, and play. This also explains why I’ve not been writing as consistently as I used to. I’ve also gained a newfound respect for people with young children who write consistently.

Built ApplyingML.com

To make learning JavaScript more fun, I decided to build something—but what?

I noticed that the applied-ml repository had high engagement, and received feedback that it was sometimes shared internally within teams—this suggested that there was demand for more practical, applied papers. I was also inspired by Vicki Boykis’ post on the ghosts in the data, especially her point on ghost knowledge. I realized that much of what we know about designing, building, and operating machine learning systems was tacit knowledge and hard to obtain other than via hands-on experience or mentorship.

Thus, I combined both ideas and built ApplyingML.com. It collects tribal knowledge on how to apply machine elating in the form of papers, tech blogs, and interviews with experienced practitioners. So far, close to two dozen ML veterans have contributed interviews. I’m quite happy with how this turned out and might write content specifically for this site next year.

Want to contribute an interview or make a suggestion? Make a pull request!

Statistics and charts for 2021

In 2021, I wrote 34 posts. The most common words are shown below, and I’d say the themes are (i) user and item data, (ii) models and features, (iii) machine learning and recommendations, and (iv) writing.

Wordcloud of my writing in 2021

I also distilled common topics and identified what resonated with people—this led to a new start-here page. The bulk of it is around machine learning systems, machine learning in production, data science methodology, writing, learning, and career.

Total page views grew from 260k to 337k (+30%) though traffic is spiky. A big chunk came from two pieces that trended on Hacker News: You Don’t Need Another MOOC and The First Rule of Machine Learning.

Site traffic from the start of the year

Here are the top pages in 2021.

Top pages on my site in 2021

Some favorites and what I think are most useful this year are:

• The Why, What, How Framework for Writing
• Design Docs for Machine Learning Systems and the template
• Metagame for Applying Machine Learning (which led to ApplyingML.com)
• Influencing without Authority
• Teardowns on recsys, search, etc.

Clicks via Google Search increased from 18k to 50k (roughly 2.5x) while impressions increased from 388k to 1.5M (nearly 4x). The reduced “efficiency” is reflected in the CTR reduction from 4.8% to 3.3%. I haven’t spent any time on SEO and it’s still low priority.

Google Search performance in 2021

Email subscribers grew from under 800 subscribers to 2.5k. Twitter followers grew from 2.2k to 6.5k while LinkedIn followers grew from 4.3k to 8.8k.

• • •

That’s it for my 2021 year in review. Hope you had a fruitful year as well. If you’re writing a year in review, please send it my way! Love to hear about how you’re growing and have been over the years.

If you found this useful, please cite this write-up as:

Yan, Ziyou. (Nov 2021). 2021 Year in Review. eugeneyan.com. https://eugeneyan.com/writing/2021-year-in-review/.

or

@article{yan2021review,
  title   = {2021 Year in Review},
  author  = {Yan, Ziyou},
  journal = {eugeneyan.com},
  year    = {2021},
  month   = {Nov},
  url     = {https://eugeneyan.com/writing/2021-year-in-review/}
}