Scaling ML at ASOS: My Journey to Senior ML Engineer
Decoding the 5-stage interview process and the engineering philosophy behind fashion tech at scale.
Fashion and technology move incredibly fast. At ASOS, that velocity is driven by millions of global customers and a “North Star” vision for data and AI moving beyond just “doing ML” to building a resilient, scalable platform that empowers entire product squads.
After a rigorous 5-stage interview process, I am thrilled to share that I have joined ASOS as a Senior Machine Learning Engineer. In this post, I am pulling back the curtain on the interview journey and the systems-thinking required to build MLOps at this scale.
The Interview Gauntlet: A Technical & Cultural Deep-Dive
1. The Initial Screening: Defining the Vision
The process began with a conversation focused on alignment. We didn’t just discuss my CV; we discussed the “Platform vs. Product” challenge. The key theme here was how to transition from fragmented, siloed team workflows to a unified, centralised MLOps model. This was my first opportunity to discuss the measurable impact of my previous work, specifically the 40% reduction in deployment time, and how that translates to the ASOS scale.
2. The Technical Deep-Dive: Template Modularity
With the Tech Leads, we went into the weeds of “Template Philosophy.” We discussed:
Infrastructure as Code (IaC): Moving from manual setups to Terraform-driven, repeatable environment management.
Governance & Security: Implementing RBAC, Service Principals, and Managed Identities within private VNETs to ensure data security without creating developer bottlenecks.
The “Modular” Approach: How to build templates where teams can swap in their training scripts while keeping the deployment, registration, and monitoring hooks standardised.
3. The Coding Challenge: Fundamentals First
This was a challenging, practical round. I was asked to implement a Linear Regression problem without using scikit-learn.
The Goal: This wasn’t about library proficiency; it was about demonstrating a deep grasp of ML internals, gradient descent, matrix operations, and loss functions.
The Takeaway: When high-level abstractions fail, a Senior Engineer must understand the math under the hood to diagnose the problem. It was a test of engineering fundamentals, not API memorisation.
4. System Design: The Centralised “North Star”
We architected an end-to-end MLOps platform. The success of this round hinged on balancing two competing needs:
Autonomy: Letting product squads move fast.
Centralisation: Providing a model registry, automated CI/CD, and monitoring hooks that work for everyone. We specifically discussed how to integrate feature management and observability (Azure Monitor/Application Insights) to solve the “silent model degradation” problem.
5. Behavioural Round: Culture as a Skill
In the final discussion with the VP of Data and the Director of ML, the conversation shifted from what I know to how I work. We discussed:
Challenging with Kindness: How to influence stakeholders and product teams when there is a disagreement on technical debt or validation standards.
The “No Ego” Culture: The team’s emphasis on being “humble, collaborative, and easy to work with.”
Ownership: Describing end-to-end project ownership—from the initial “we have a problem” to the final adoption metrics.
Key Takeaways for Senior Engineers
Influence is a Technical Skill: Your ability to move a team to a new standard (like a shared pipeline) is just as critical as the code you write to build it.
“No Ego” is a Multiplier: The best engineers focus on building shared solutions, not just “quick fixes.” Being approachable and willing to mentor is what separates “good” from “great.”
Master the Basics: Whether it’s writing algorithms from scratch or understanding cloud security networking, don’t let tools replace your fundamental understanding of the system.
I am incredibly excited to start this chapter with the ASOS team, helping build the platform that defines the future of fashion tech. If you are an engineer working on MLOps infrastructure, I’d love to hear your thoughts on how you balance platform governance with team autonomy.
Prasad Zende
Senior Machine Learning Engineer @ ASOS