Full disclosure, I work at Datadog as a research engineer and this was my team behind this.
We’re releasing an open-weights time series foundation model and 350 million datapoints as a benchmark (Apache 2.0 license). The model and dataset are up on Huggingface, the paper is on Arxiv (links in the blog post).
The model and dataset all come from our own internal telemetry. We dogfood extensively and pride ourselves on being best in the world with dog puns.
I joined the team 10 months ago, coming from a data engineering background. This was my first research project. The amount of work that went into this paper blew me away. It was scary taking on a new role mid-career like this but very rewarding, I learned a ton moving from data engineer to research engineer.
Give the model a try, grab the dataset, and let us know how we can help.
- don’t plan too far into the future, the further you do the less specific you should be
I like annual visions for each area of my life, so once a year reflect on each area and think where you’d like to be a year from now. My annual time capsule I call it, is coming up next week and I can’t wait!
- do monthly reflections
I have a google calendar reminder that links to a journal template. Here I briefly reflect once a month on what’s working and what’s not in each area
This is more tactical. Habits and things to help implement the vision
- ad hoc reflection
When struggling with something or needing to evaluate, I’ll just open a doc and type. Stream of conscious style to get it all out there.
Then read it and pluck out next steps. Sometimes this means pivoting, sometimes it’s just a deeper understanding of why the path I’m on is the right one.
Sorry for the sloppy writing here. I actually would like to write this up better when I’m not on mobile.
I'll follow on here because "don't plan too far in the future" is something I struggle with. My imagination is good, and so staying away from 'what could be' can be difficult at times.
---
My "plans" typically start with a broad vision. These ideas/visions come to me in day dreams as I ponder current tech or things that are interesting to me presently. These visions of what could be are typically not very revolutionary, and mostly just what ifs. Recently I've been pondering geographically bound search engines aimed at getting people back into the real world (I think this is a product in search of customers, tbh.), or a language server and dev tools for an esoteric scripting language, or a HN clone that serves specifically to catalog AI stuff.
From a broad vision, I make a small bulleted list of what a finished MVP might look like.
After the MVP I make/stumble-into an assessment of my skills and the skills required to complete the project. In my current project -- the aforementioned HN clone -- I found that I knew almost nothing about ASPNET web projects, then I found that I knew nothing about SPA+ASPNET, then I found I knew nothing about user authentication/authorization, and so on. This step is "pay down your debt of ignorance". It's kind of grindy, but I've been enjoying learning as I go. I've restarted my coding project about 10 times as I attempt to reformulate based on some new thing I've learned.
And then wash, rinse, repeat until I check all the boxes for the MVP. Then deploy and advertise/tell people about it.
If you're sick -> take care of that first. Don't work.
If you're just having an off day, trouble focusing, not feeling in the mood, that's different. I can suggest a few things that's worked for me.
1) Focus on just doing one thing.
Pick your most important project, then pick one specific piece of that, and shut down everything else until you at least complete that.
2) Set a timer.
If you're problem is getting started on a new or challenging project, do something like the pomodoro technique where you just promise yourself you'll work on it for X minutes and then stop. This removes some of the barrier to entry and can help you get started.
3) Remove distractions.
Similar to the first point, but is Slack always on? Are you getting notifications from your phone? Checking e-mail? This is a problem with our current life, we have too much competing for our attention. A lot of what we need to do to be successful is blocking out the noise so we can focus.
I like the distinction. Yesterday was a mix of days 1 & 2. Agree with all suggestions (I have a quiet desk with notifications turned off usually, so I guess that wasn't the problem).
I'll implement the just focus on one part of one project thing though, sounds like a smart strategy.
I graduated college absolutely certain I would be a day trader on Wall Street for the rest of my life.
And I did that for about 7 years.
But 5 years in or so I started to realize I didn’t quite like it. I actually was more interested in writing code to solve my trading problems. But what really appealed to me there was the writing code bit.
It took about a full year of my mini identity crisis for me to realize, there were other jobs out there I could enjoy. And the career my 20 year old self had picked wasn’t my ultimate path.
It’s OK to change directions.
Continue to reflect on what you want for yourself. Don’t try to predict too far out because it will likely change. But think about the next year or three, where do you want to be?
Once you have a rough idea in mind you can think backwards about what steps are needed to get there.
Hang in there, self reflection is tough but very worthwhile.
Your previous experience absolutely has and will NOT go to waste. I can’t tell you how many times my finance experience has helped me at my software jobs or just in life.
While it may not be directly related to your new job, that experience is there, helping you in more ways than you realize.
We’re releasing an open-weights time series foundation model and 350 million datapoints as a benchmark (Apache 2.0 license). The model and dataset are up on Huggingface, the paper is on Arxiv (links in the blog post).
The model and dataset all come from our own internal telemetry. We dogfood extensively and pride ourselves on being best in the world with dog puns.
I joined the team 10 months ago, coming from a data engineering background. This was my first research project. The amount of work that went into this paper blew me away. It was scary taking on a new role mid-career like this but very rewarding, I learned a ton moving from data engineer to research engineer.
Give the model a try, grab the dataset, and let us know how we can help.
We are excited to see what you build!