Software Engineer (Machine Learning) New UK (London, Manchester, Bath)

Greenhouse Software, Inc.; Greater London, Manchester, Bath; Hybrid; GBP 70,000 - 110,000; Full time

Job summary

Netcraft is seeking an ML engineer to develop adversarial classifiers that block cyber threats. You will deploy models on GPU-backed Kubernetes clusters, integrate vLLM-based inference, and collaborate with engineers and analysts to improve detection at scale.

You’ll train and evaluate models, run experiments, and monitor performance in production, adapting approaches as attack patterns evolve. The role emphasizes pragmatic ML and reliable systems in a security-focused environment.

Qualifications

• Proficiency with web fundamentals (HTTP, DNS, how the web works).
• Understanding of ML training, evaluation, and precision/recall concepts.
• Strong written and verbal communication; ability to articulate impact clearly.
• Empirical mindset: design experiments, measure outcomes, use metrics.
• Willingness to learn Go and Perl (async programming).
• Attention to detail and accountability for production classifications.

Responsibilities

• Build, evaluate, and tune classifiers for blocking decisions.
• Deploy and operate model serving infra on Kubernetes, including LLM inference.
• Optimize models for high-volume, low-cost serving.
• Run experiments on live data; measure precision, recall, false positives.
• Integrate models into production classification platforms.
• Create and maintain pipelines for features, retraining, and metrics.

Skills

Web fundamentals

ML fundamentals

Communication skills

Empirical mindset

Go & Perl learning

Attention to detail

Tools

Kubernetes

Flux

Argo

API gateways

LLM inference (vLLM)

MLflow

Argo Workflows

llama.cpp

SQL databases

Job description

Netcraft is the global leader in cybercrime detection and disruption. We’re a trusted partner for three of the four largest companies in the world and many large governments. We've blocked over 225 million cyber-attacks to date, and we take down around 33% of the world's phishing attacks.

Our purpose and passion are focused on just one thing: protecting the world from cybercrime.

That passion shapes how we work, too. We’re proud of our talented team and the value each person brings, and we’ve built a workplace where people feel supported and inspired. From strong benefits and wellness programs to meaningful collaboration and team connection.

The role

We are looking for an engineer who wants to apply ML to a real adversarial problem, where attackers actively evolve to evade you and turn to increasingly complicated schemes. As part of the Novel Classification team, you’ll use machine learning and AI to identify fraudulent content on the Internet, from phishing sites to the full breadth of online scams.

You’ll learn from experienced engineers and see your ideas in production within weeks, driving the automated blocking and takedowns that disrupt cybercrime at scale. Recent projects include multi-modal embeddings that power classification and search, LLM classifiers, computer vision models for logo detection, and ensemble predictors that combine visual and semantic signals. We keep a close eye on the field and adopt new models and techniques early when they look promising.

You’ll work across the full lifecycle: training and evaluating models, serving them on our GPU-backed Kubernetes clusters, wiring them into production classification systems, and watching how they behave against live attack traffic.

Day to day, you’ll be:

• Building, evaluating and tuning the classifiers that decide what gets blocked and taken down.
• Deploying and operating model serving infrastructure on Kubernetes, including LLM inference (e.g. vLLM).
• Optimising models for high-volume, low-cost serving.
• Running experiments against live data, measuring precision, recall and false positive rates, and using the results to decide what ships.
• Integrating models into our production classification platforms.
• Building the pipelines and monitoring behind the models: feature extraction, scheduled retraining, experiment tracking, and the metrics that prove a solution works.
• Investigating and fixing issues in production, from model regressions to performance bottlenecks.
• Working across the business with engineers and analysts who depend on our classifiers.

What you’ll need to be successful

• Strong web fundamentals: HTTP, DNS, and above all how websites work, from request to rendered page. You'll use them both to understand the resources we classify and to build high-performance webservers.
• A working understanding of machine learning fundamentals, such as training, evaluation, overfitting and precision/recall trade-offs. This could come from industry, research or substantial personal projects.
• Good written and verbal communication skills. The work the team does is complex, and we must be able to articulate the impact of our work clearly and concisely, to a wide range of stakeholders.
• An empirical mindset. You design experiments, measure outcomes, and use metrics to validate and demonstrate impact.
• Willingness to learn Go and modern Perl, including asynchronous programming.
• Strong attention to detail and a real sense of responsibility. Customers act on our classifications, and Netcraft's reputation depends on getting them right.

Bonus points if you have

• A keen interest in cybercrime or cybersecurity.
• Experience fine-tuning models, including evaluation.
• Experience with LLM agents, evaluation & serving (e.g. vLLM or llama.cpp).
• Familiarity with cloud-native technologies such as Kubernetes, Flux, Argo and API gateways.
• Experience with multi-modal models, embeddings or similarity search at scale.
• Experience with ML experiment tracking and workflow tooling, such as MLflow or Argo Workflows.
• Experience with SQL databases, including writing and optimising queries against large datasets.

Preferably based in London, though Manchester and Bath are also options.

AtNetcraftwe strongly believe in individual growth and development. Even if youcan'ttick every box above right now, pleasedon'tlet it put you off applying.

The Reward Package:

• 33 days annual leave - includes public holidays which can be worked if you prefer
• Generous enhanced family leave and pay
• Enhanced company sick pay from day one of employment, with additional support after your first six months’ service
• Salary sacrifice pension with matched employer contributions up to 5% + NI savings
• Private health cover, including access to a private GP service
• Life Assurance at 4 x salary
• Equity tracking scheme (eligibility criteria apply) - allowing you share in Netcraft's long term success
• Hybrid and flexible working options
• Meals, snacks, and drinks provided in the office
• Comprehensive health, safety, and wellness service available 24/7 for you and your family
• Two paid Learning & Development days per year plus access to learning resources such as Udemy and Coursera to support your career
• Two Volunteering Days per year, to allow you to contribute to your wider community or charity
• Spot reward bonus scheme to reward truly outstanding contributions

Diversity, Equity & Inclusion

This is deeply important to us. Through our ally network, we support under-represented groups andmaintaina working environment free from bias, harassment, or discrimination. We encourage candidates from all backgrounds to apply.

We’re also happy to make adjustments to the hiring process to ensure every candidate can participate fully.

Please note: Netcraft does not accept unsolicited approaches from external recruiters.