My First 30 Days at Ankar: Autonomy, Debugging, Deadlines, and Discovery

“My first 30 days at Ankar: writing code, breaking → fixing → building things, learning about real-world AI systems at a breakneck pace”

When I joined Ankar, I expected to build cool machine learning systems that could make sense of innovation - research data, patents, office actions and such. What I didn’t realise was that the real experiment would be me; learning, adapting, and rethinking how I build as an engineer, all in just 30 days.

Ankar’s mission is bold: building AI that helps people understand innovation itself. But what really makes the company stand out isn’t just the technical challenge; it’s the culture of trust, speed, and ownership that pushes you to do your best work.

Here’s how my first month went.

Day 1 - Orientation (8th September 2025)

The first day felt like a breath of fresh air. The founders welcomed the new hires, and one line really stuck with me, which I found really refreshing: “It’s not only okay, but expected to challenge our views. The best idea always wins”

It was clear from the get-go that Ankar wasn’t a place for hierarchy - it was a place for ideas. The culture seems to be everything here at Ankar; of course, provided you're talented enough to make the cut.

Day 2 - Local Dev Setup (9th September 2025)

Nothing like starting your journey with a little debugging marathon. I spent the day piecing together how the backend, ML service, and frontend interacted with each other. 

While doing so, I chanced upon a sneaky bug. Mistakes make for stronger learning signals than wins in my opinion, so understanding the bug became my task for the day. Went stack trace diving with Nick (Head of Engineering). He’s a solid mix of deep technical understanding of the product, and an extremely approachable, collaborative style of working. Fun fact: during my interview, one of the engineers told me that Nick was “the rare kind of engineering lead who still codes and still cares.” They were right, which was great for me as that was high on my list of priorities when I was choosing where to work.

The dev talent here? Smart, hard-working, and relentlessly helpful… the trifecta of a solidly curated dev team.

Day 3 - First Real PR (10th September 2025)

Squashing bugs ahoy! The bugs I discovered yesterday? Got my baby PR with changes to them merged. Getting accustomed to the stack, and kindling a new-found interest in software engineering. For context, we ML Engineers typically spend a healthy portion of our time staring at data, so the crash course in good dev practices from a team of, well, good devs, was very helpful.

Ended up doing several review cycles and learning Ankar’s process the hard way: tickets, context, and code hygiene matter.

We also had an office coffee taste test to see what everyone liked best. I was the only one who voted for instant. Contrarianism is a core cultural value here… right?

Day 4 - My First Sprint Planning! (11th September 2025)

Joined the Office Actions team and got my first taste of sprint planning. Nick reviewed my earlier backend PR, and by the end of the week, it was nearly ready to merge. Slowly starting to find my rhythm.

Day 7 - Learning the Ankar Way (14th September 2025)

Had my second 1:1 with Nick. He told me something that stuck: “Within six-ish weeks, you should be operating autonomously - minimal support, full ownership.”

Challenge Accepted. Doable, I think. He believes that I'm performing exactly at the level I need to be right now.

He also broke down the three pillars of thriving at Ankar:

• 1) Ownership - Don’t just do the task, make sure it’s done

• 2) Speed - Ship faster, iterate faster

• 3) Communication - Keep everyone in the loop, that’s how a start-up scales

Day 8 - Action Suggestions (15th September 2025)

Started working on claim extraction improvements and iterated on examples for better model performance. Waiting on review.

Day 15 - Merge Fatigue (22nd September 2025)

The past one week has been hectic in working on my new PR. I’m now in the final stages of getting my new PR merged - just having to make sure that it passes the design and coding standards of the team.

This PR required several back-and-forths between me and my reviewer, and I'm really grateful for them. The process was exhausting but invaluable. Every comment pushed me to think more systematically, more like a builder than a coder.

I’m realizing that the skills I’m sharpening here - debugging complex systems, reading design decisions deeply - are going to pay dividends long-term in ML engineering.

One thing I have realised is that the pace here is fast. Real fast. We iterate quickly.
Speed doesn’t mean chaos here, it means clarity and ownership.

At Ankar, you don’t wait six months to ship something users touch. You do it in week two. That’s the kind of feedback loop that rewires you as an engineer.

Day 18 - Deadline Approaching [iterate fast] (25th September 2025)

Wiem (co-founder), dropped a message in the Office Actions slack channel. A client wants to use the tool for US patent cases. Fast. Cue the chaos (the good kind, the type you would want in a start-up).

I dove into the task, which spanned both the ML and backend stack. Pro-tip: workflow engineering (shortcuts, window management, tooling, even split-keyboards :p ) are lifesavers.Overall, a rewarding experience.

It was one of those “this is why I do this” moments. When discomfort turns into growth. Putting myself in unfamiliar situations and having to stretch myself mentally is quite a powerful stimulus for improvement

Day 21 - First Open-Ended Task (28th September 2025)

It’s now midnight. I’m working on document column highlighting and reviewing some feedback from Oliver, one of our backend devs.

After I got home, I fell down a YouTube rabbit hole and ended up watching Sylvan Franklin’s video on how to improve 10x faster than 99% of engineers. In the comments section, there was a specific comment that stopped me mid-scroll (by user @astrixx)

The comment said

That comment hit home. It’s basically the Ankar ethos in one paragraph: work with people who challenge you, stay curious about the “why,” and never settle for just working code.

Day 30 - 1 Month In (Reflections and Such)

It’s already day 30. The last week has been hectic again - in the best possible way. I’ve been deep in my new PR and have been given ownership of the ML service behind our Office Actions tool. Time really does fly when you’re learning at an accelerated pace. It’s only been a month, but it feels like I’ve grown the equivalent of six.

Reflecting on the past month, a few things stand out:

• Massive autonomy + ownership: You are trusted early and expected to deliver - it’s daunting, but it’s the best kind of pressure. Being able to shape the product from your early days is mildly daunting, but more exciting

• High trust, high velocity: Slack replies at 1am are common. Not because of overwork, but because people care. Expect to spend a portion of your weekend thinking about how you can make your code better, more performant, or efficient

• Ask questions: The fastest way to learn here is to ask questions. Don’t be shy, put aside your ego, and clarify everything. People here are willing to help you and push you to be a better version of yourself

• Enjoy the ride: Whether it's engaging in office banter, the monthly dinners, the Ankar Cup challenges, or exploring the vibrant Hoxton food scene via an in-house curated food guide

Looking Ahead

The next 30 days will bring more PRs, more late-night debugging, more refactors, and more caffeine. But if the first month taught me anything, it’s that Ankar doesn’t just hire engineers - it builds them. 

Ankar has been the fastest environment I’ve ever found for growing as an engineer. You’re trusted early, challenged often, and surrounded by people who care about helping you become world-class. If you’re the kind of engineer who learns by doing, thrives under real ownership, and wants to see your work make impact fast - you’ll probably love it here.

You don’t join Ankar to be managed.
You join to level up.

And that’s something worth debugging for.

—

This article is written by Manick, our Machine Learning Engineer. He holds a MSc in IoT and a Bachelors in Mechanical Engineering; and previously spent 3 years as a SWE and 4 years as a MLE