The Anti-Stealth Startup
The inner workings of building an open-source startup.
We already are up to 41 subscribers after the anti-stealth post got a bit of lift-off on LinkedIn. I'd like to officially welcome every one of you into the ZenAllies herd (That is our internal name for what we call you guys, i.e., the people who will follow us on this open journey).
So without further ado, here is the second edition of "Building an MLOps Startup in Public"
👀 What is this newsletter going to be about exactly?
For those of you who don't exactly know who I am, my name is Hamza Tahir, and I'm one of the co-founders of ZenML. We're a company based in Munich, creating an open-source MLOps pipeline framework. The vision here is to bring together the noisy MLOps landscape, and empower the producer of the ML models (i.e. the data scientist) to take ownership of their ML models to production.
As outlined in the first edition of the newsletter, we're a bit sick of this notion of being a stealth startup. We want to be able to share this weird-ass startup journey with everyone, completely unfiltered. Then we can make mistakes, learn, and grow. Also, it helps us personally to just document this journey from the early stages. Please expect a large amount of information that is hopefully insightful into how the team at ZenML is approaching building this company.
We'll try and be an open book in this newsletter: stuff you don't usually share from the trenches. The goal is to give an inside look into how a startup behaves and acts, hopefully inspiring other people to set out on their own journeys. It really is just a bunch of people with a shared idea. Let's try and pull the curtain away from the hype shall we.
If you happen to like this newsletter, and know someone else who might like it as well, then feel free to spread the word 📢—the number of subscribers to this newsletter is a key KPI we're measuring!
👪 The family is growing
The ZenML team just went from 3→9 people. 5 of us are already on board (see everyone here). By January, the rest will join. I'd like to just introduce them quickly:
Alex Strick van Linschoten: Based in London, Alex has the most unique story of transitioning from a history research background to hard-tech!
Michael Schuster: When not practicing being an excellent engineer, Michael can be found hanging out in remote AirBnb's in Austria, looking for places to ski.
Baris Can Durak: Baris plays assist in both ML Engineering and DOTA. He is the nicest person around, and can solve all your caffeine requirements.
Adam Probst: Adam is a master of connecting the dots, and loves the art of engaging with people from different backgrounds.
The rest I'll introduce next time! They're all very cool—trust me.
P.S. We're not all based in Munich → The picture above was a few days ago when we managed to get a bunch of us together. So we have a semi-remote working culture!
🛫 ZenShare and ZenHack
We just initiated two internal events called
ZenHack and had their inaugural meetings.
ZenSharesare meetings where one developer shares code with other developers to showcase cool stuff they implemented. It's like a joint code review / code brag meeting we use to learn and share. There is not a schedule for it, but we do it every once in a while when something really cool pops up.
ZenHacksare internal hackathons we're having where we put ourselves in the shoes of the user we're building ZenML for. We take a cool MLOps problem and try to solve it within a day using ZenML. We plan to do this monthly, and publish blogs and examples based off of the results. It's a great way to build empathy for our users and find weird bugs in the developer experience.
Both have already been super fun and we hope to continue the tradition. We'll be posting to our blog about the ZenHacks whenever they happen, so watch this space!
📈How we're measuring traction
Open-source traction is notoriously hard to measure and there are so many resources on it, from the Linux Foundation to the Venture Capitalist angle on it. We're currently still finalizing our so called 'North Star' metric but collecting a lot of data to make that decision.
We collect data via a daily scheduled Google Cloud Function that runs a Python script which collects various sources, such as and then have a GCP function that runs every day to collect data from various sources (website visits, Github stars, contributors, views etc). This data is stored in a Notion database and additionally a BigQuery table. We might open-source this function later, its actually quite cool → Reply directly to me here if you think we should!
Every Monday, the function posts on our internal Discord channel, with some emojis showing how we compared to last week:
This gives full transparency into the impact work is having and seeing those little 🔥 emojis gets everyone fired up after the Monday standups!
🎉 Releases: ZenML 0.5.0—ZenML 0.5.2
I don't really want this newsletter to be a lot about the product, but it still makes sense to give everyone an update on the major highlights. I mean the product is the main thing happening in the company right?
We took around 2-3 months to completely revamp the internals of ZenML. We had fresh blood and ideas coming in, and we knew we needed a clean start to build on. We are now finally at the end of this ramp-up period with the post 0.5.0 releases, and it is such a relief. We knew our small pre-seed community was slowly fading away and it felt like a race to get to the refactor done and not lose all momentum. But it feels tremendously nice to see the new ZenML, and feels like it was the right decision. We will know in the future whether it was or not of course.
Why not give it a spin?
pip install zenml
and then visit the brand-new step-by-step guide in our new docs.
Or just run through the quickstart colab for an even quicker look.
Now we return back to getting eyes on the product—which of course means content content content. Which brings me to...
We had this idea in the company that we need a character that represents our target user. After much brainstorming, and deliberating, we commissioned the brilliant Diana Nezhivova to create Sam—the data scientist who just wants to make models work in production:
This is just the start of Sam’s story to find Zen in her ML—expect a lot more to come from her!
✍️New Blog, New Content!
We also revamped our blog!
Alex Strick van Linschoten wrote about how to level up your testing with Python: It's a great read for those looking to level up their testing game with Python.
I wrote a general vision piece about how ZenML is taking on the ML pipeline challenge: It goes a bit in depth on the ownership dilemma for data scientists in production.
Do take a look at the blog: it actually has a lot of content that you might find interesting if you're interested in MLOps.
📢 The blog is just one part of our outreach and content strategy (of which this very newsletter is a part of of course). We'll write more about this in the upcoming newsletters.
Some of you might have noticed all the team are male. Yeah, that is definitely not the intention while hiring. We are actually very proud of how diverse the team is in terms of background, skill set, and personality. But currently we're still looking to hire our first female talent. We've made some efforts towards this direction and fingers crossed we find a stellar candidate! Any advice is welcome of course.
There are a lot of moving pieces with building this startup, which are so hard to manage. Here is a sample:
Keeping the public facing stuff synced with a fast moving and evolving product: Website, Docs, README etc.
Prioritizing features and trying to understand what to build next.
Trying to revive the community (After our initial break we've had to sort of restart)
Hiring... It takes a lot of time but is extremely interesting.
Keeping investors happy and to make sure they don't get nervous (we love you all btw 😚)
Coordinating internally all of this and having fun along the way!
❓Questions for the community
There is so much going on and its hard not to make this a
spam diary. How much more/less insight would you like into the company? Anything you're particularly interested in?
Are you (potential) users of ZenML or just interested in the startup journey?
Would you be willing to share this newsletter? If yes, it would help us out a lot!
Alright, this turned into one giant newsletter, so I'll stop here for now. Looking forward to hearing from some of you 41 people (just drop a hi to connect if you'd like, we're small enough that that works out). Thank you, and wish you an amazing week🌞!-
P.S. Original ZenAllies, I added you to this list as it just made sense to not have two newsletters. Please opt-out if you don't like it!