AI Engineer

Department:Eng Growth & Integrations 

Employment type: Full-time

Location: UK, Spain, Bulgaria, Germany 

WebPros is committed to empowering businesses worldwide through cutting-edge solutions in web hosting, billing automation, infrastructure, server management, and online marketing. Since our founding in 2017, we’ve rapidly grown into a global leader, expanding our robust portfolio to include industry-defining brands such as cPanel & WHM, Plesk, WHMCS, SolusVM, XOVI, SocialBee, Sitejet and Comet Backup.

Today, we power 85 million+ websites across 900,000+ servers worldwide, backed by a 650+ strong team of dedicated professionals spanning multiple continents. At WebPros, we embrace a fully remote, borderless workplace with a focus on flexibility, autonomy, and work-life balance. Whether you prefer working from a bustling city hub or a quiet home office, we enable our teams to thrive from anywhere while staying deeply connected through our strong collaborative culture.

Hi, I'm Deniz Okcu, Engineering Manager at WebPros, and your future manager.

I run the engineering teams behind our growth products, the tools small businesses and agencies use to get found online. For twenty years that meant Google. Now it increasingly means ChatGPT, Perplexity and the AI summaries in search results, and almost nobody has worked out yet what "ranking well" means in that world.

RIVA is our answer to it. It looks at a business the way a language model would, works out how the model understands and describes it, then tells the customer what to change. It ships alongside XOVI's established SEO suite, so there are real customers on the other side of what you build rather than a prototype hunting for a market.

For the role:

You would own the engine: the analysis algorithms, the retrieval, the evaluation harness, and the agents that turn a finding into an applied fix. Not prompt-tweaking on top of someone else's model, but the logic that decides whether what we tell a customer is true.

That last part is really the job. Anyone can get a language model to produce a plausible recommendation. Proving it is correct is the hard part, because the ground truth shifts every time a model is updated, and it is the reason a customer would pay us rather than ask ChatGPT themselves. If you've ever been irritated by AI tools that sound confident and can't show their work, this is the job where you get to fix that.

• You own the analysis engine. The attributes and scoring rules that decide whether a model can understand a business are the core of the product, and keeping them at their sharpest is on you. When model behaviour moves, you're the one who spots it, works out what it means for our rules, and either ships the change or makes the case for it.

• The evaluation framework. Ground-truth sets, metrics, and regression detection that survives the fact that the thing you're measuring answers differently every time you run it. Most of the rest of this list is blocked without it.

• Retrieval. Getting text out of real websites, deciding how to chunk it, picking an embedding model, and keeping a path from every claim we make back to the page it came from.

• The agentic architecture. One agent, several, or a plain tool-use pipeline. You'd work out which fits and say why in terms of cost and latency rather than preference. We haven't settled this yet.

• Model selection and cost per run. Benchmarking frontier models against open-weight ones, then routing so we only pay for the expensive option where it changes the answer.

• Guardrails and observability. Validating what goes in and what comes out, resisting prompt injection, catching hallucinations, and enough tracing to know where the money and the latency actually go.

• Turning research into production code. Reading what gets published on LLM and GEO behaviour, and making the case with evidence when it says we should change approach.

Must-haves 

 

• Production Python. It's our stack and you'll be in it daily, alongside FastAPI, PostgreSQL and pytest.

• You've used the LLM SDKs directly rather than only through a wrapper, and you know the point where an orchestration framework like LangChain, LangGraph or Pydantic AI stops paying for itself.

• You've built retrieval that runs in production, including the messy parts: extraction, chunking, choosing an embedding model, recognising when retrieval isn't the answer, and a vector store such as pgvector or Qdrant.

• You've evaluated an LLM system properly. A real dataset, metrics beyond exact match, and a judge you checked against human labels rather than trusted. Langfuse, Promptfoo and Ragas are the sort of tooling we mean.

• You understand model behaviour well enough to be sceptical of it. You can form a hypothesis about why retrieval or ranking behaved the way it did, test it, and say plainly what you found.

• Prompt, context and harness engineering. Writing the instruction is the easy part. What counts is what goes into the context window, what you leave out, and what you build around the model. DSPy is one example of the tooling here.

• You've designed agentic systems and can say why you picked that architecture over the ones you didn't.

• You've kept one running in production. Tracing, token and cost attribution, latency, error rates, and using those numbers to decide what to fix first. Langfuse, LangSmith and Prometheus sit in this space.

• You know how these systems fail adversarially: prompt injection through content you crawled, output that reads well and is false, and what validation actually catches. Guardrails AI, LLM Guard and NeMo Guardrails are examples.

Nice-to-haves 

• Familiarity with SEO, generative engine optimisation (GEO), or search-visibility products.

• Classical NLP: classification, named entity recognition, sentiment analysis.

• WordPress, CMS integrations, schema.org, or llms.txt.

• Crawling and extraction at scale, including JavaScript rendering, rate limits and content quality.

• AI governance and risk management practice, such as the EU AI Act.

None of these are dealbreakers, and you don't need an SEO background because we have that in house. The LLM depth is the part we can't substitute.

30 Days — Learn & Integrate 

 

• Get set up, get into the RIVA codebase, and follow one analysis end to end.

• Learn our current attributes, scoring rules and visibility metrics (Mention Rate, Average Position, Query Coverage) well enough to explain them back to us, and to tell us which ones look shaky.

• Meet the product, engineering and XOVI SEO people you'll be working with across time zones.

• Ship a first small improvement to the pipeline, mostly so you've been through the whole path once.

60 Days — Contribute & Collaborate 

 

• Deliver a first working evaluation harness with a ground-truth set, so we can finally argue about quality using numbers.

• Correct or extend a set of scoring rules on your own, with evidence for why the new version is better.

• Be active in code review, giving feedback as readily as you take it.

90 Days — Take Ownership 

 

• Ship a RAG pipeline that produces grounded, content-specific suggestions a customer can act on directly.

• Measurably increase how deep and how widely we analyse a site.

• Own a core area of the analysis engine end to end, without needing to check in on it.

120 Days — Drive Growth 

 

• Set how we validate algorithm changes, so quality holds as the team grows around you.

• Bring us your own read on where models are heading and what it means for our rules, without waiting to be asked.

• Lead the design of the next big capability, improvement agents or sentiment analysis, and get it into the roadmap.

• Spread what you know across engineering and product, so the AI expertise isn't sitting in one head.

• You want a spec and a definition of done handed to you.

• Shifting priorities wear you down. Models move, and this roadmap follows them.

• You'd rather implement a decision than be in the room making it.

• You don't enjoy having your conclusions challenged, or challenging someone else's.

• You need people in a room to do your best thinking.

• Team & Workflow: Remote and asynchronous, across several European time zones.

• Collaboration & Ownership: Self-managed work, with a strong emphasis on being autonomous and proactive with good time management.

• Tech Stack & Tools: Python 3.12, FastAPI, PostgreSQL, LangChain with the OpenAI and Anthropic SDKs, Langfuse for tracing and evaluation, DSPy for prompt optimisation, BeautifulSoup and lxml for content extraction, Docker and Kubernetes, pytest and mypy. Git, Jira and Confluence around it. Agent orchestration and the vector layer are still open choices (LangGraph, Pydantic AI, or our own control flow; pgvector, Qdrant or similar), and whoever takes this role gets to make them.

• Testing & Quality: Code review as standard, plus evaluation-driven development. Algorithm changes get validated against ground-truth data and checked for regressions before they ship. We'd rather delay a change than ship one we can't measure.

• Releases: Things go out as they're ready, not on a quarterly ceremony.

 

How you'll work (AI-native)

 We work AI-native and expect the same here: 

• You practise agentic engineering: a structured process where intent is the source of truth, and where you review what the agents produce carefully enough to know it's correct and well designed.

• You know the building blocks of harness engineering (MCP, skills and plugins, subagents, spec-driven development frameworks) and which to reach for when.

• You'd be working on a production service with paying customers, so software engineering judgement matters as much as it always did.

 

WebPros is built on diversity, not just in principle but by design. We’ve grown through acquiring industry-leading brands, bringing together teams from 42+ nationalities, 20+ countries, and 10+ brands. Instead of enforcing a one-size-fits-all culture, we embrace unique perspectives, different working styles, and localized expertise to drive global innovation.

We are a fully remote, async-friendly company where transparency, open communication, and collaborative problem-solving define how we work. Whether you thrive in the fast pace of a startup or the stability of an established product, WebPros offers a dynamic, evolving career journey that you can shape based on your strengths.

We are committed to fostering an inclusive, equitable workplace where every team member feels valued, heard, and supported because our diversity is our strength.

We believe benefits go beyond just perks. At WebPros, we invest in your growth, flexibility, and well-being, so you can do your best work and build a meaningful career. While we strive to provide equal global benefits for all team members, some may vary by location, and we’ll clarify the details with you during the interview process.

•    Competitive salary

•    Remote-first flexibility: Work from anywhere in Europe (or hybrid in available offices).

•    Development Budget: Annual personal growth allowance.

•    Learning Support: Udemy subscription & peer learning.

•    Public Transport & Meal Vouchers: Location-dependent benefits.

•    Paid Time Off: 20+ days of annual leave.

•    Team Culture: Regular team and on-site company events.

Hi, I am Zornitsa your Talent Partner at WebPros. I’ll guide you through the interview process, ensuring a clear, transparent, and meaningful experience.

Interview Process

• Screening Interview (30 min): An informal conversation with Zornitsa, your Talent Partner, designed to get to know you better. We’ll discuss your background, motivations, and career aspirations while providing a clear overview of WebPros, the team structure, and what you can expect throughout the hiring process. This is your chance to ask questions and ensure the role aligns with your professional goals. 

• Hiring Manager Interview (45 min): A conversation with Deniz Okcu about how you work, how you think through problems, and whether this team is the right place for you.

• Technical & Culture Interview (60 min): We put a real problem from RIVA's analysis engine in front of you and think it through together: how you'd evaluate it, where you'd expect it to break, what you'd try first. There's no take-home task and no algorithm puzzles. 

At WebPros, we want every team member to feel excited to bring their full, authentic self to work. Our strength lies in the diversity of our people, with different backgrounds, perspectives, and experiences driving innovation and making us stronger together.

We are committed to ensuring equitable opportunities, pay, and support for all employees, regardless of identity, background, or circumstances. We foster a culture of belonging, respect, and psychological safety, where everyone feels valued, heard, and empowered to contribute meaningfully and grow.

Discrimination of any kind, including sexism, racism, xenophobia, homophobia, transphobia, ableism, and ageism, has no place here. By embracing differences and championing inclusivity, we create an environment where everyone can thrive.

We are looking forward to your application!

Production Python experience with LLM SDKs, retrieval, evaluation, agentic systems, observability, prompt and context engineering, and production software development.

Key Responsibilities

• building algorithms
• evaluating systems
• designing agents

Skills & Tools

Python, FastAPI, PostgreSQL, pytest, LangChain, LangGraph, Pydantic AI, pgvector, Qdrant, Langfuse, Promptfoo, Ragas, DSPy, LangSmith, Prometheus, Guardrails AI, LLM Guard, NeMo Guardrails, WordPress, BeautifulSoup, lxml, Docker, Kubernetes, mypy, Git, Jira, Confluence, OpenAI SDK, Anthropic SDK, MCP

Job Details

• Category: Engineering
• Seniority: Senior Level
• Commitment: Full Time
• Workplace: Remote — United Kingdom or Germany or Bulgaria or Spain
• Salary: EUR 68,000 – 72,564 / year
• Languages: English

Benefits

• Generous paid time off

About WebPros

A global technology company providing web hosting, billing automation, infrastructure, server management, and online marketing solutions. — Industry: Information Technology