About this role.
Pleo is seeking a Senior Applied AI Engineer to design, build, and operate customer-facing AI features for its spend-management product. The role centers on production LLM systems, including RAG architecture, prompt and context engineering, agentic workflows, evaluation pipelines, and production observability. This engineer will partner closely with Product, Design, Data, and platform teams while owning scoped outcomes from problem discovery through deployment. The position requires strong Python engineering, practical data and retrieval expertise, and the ability to communicate AI trade-offs clearly to technical and non-technical stakeholders.
Role DNA
A quick view of the complexity, pace, ownership and collaboration implied by the job description.
Job Complexity
5/5Pace & Pressure
4/5Autonomy Level
5/5Communication Load
5/5Salary analysis
Estimated compensation compared with the broader US market for similar roles.
Core skills
Skills and capabilities most closely associated with this opportunity.
Cover letter sample
Dear Pleo Hiring Team,
I am excited to apply for the Senior Applied AI Engineer role because it combines production-grade AI engineering with direct ownership of customer value. My experience building and operating LLM-powered products, retrieval pipelines, evaluation frameworks, and reliable APIs aligns strongly with Pleo's goal of taking AI features from prototype to measurable product impact.
I would bring a pragmatic approach to RAG design, prompt and context management, observability, and quality measurement, while partnering closely with Product, Design, Data, and platform engineering teams. I am particularly motivated by the opportunity to use Pleo's unique spend data to solve real customer problems rather than pursue AI research in isolation.
I would welcome the opportunity to discuss how I can help Pleo ship trustworthy, useful AI capabilities and contribute technical leadership across the team.
Sample interview questions
I would begin with the user decisions the assistant must support and define success metrics such as grounded-answer accuracy, retrieval recall, latency, and escalation rate. I would build an ingestion and normalization pipeline, select chunking based on document structure and query patterns, use hybrid retrieval with metadata filters and re-ranking, and require citations or grounded responses. I would then establish offline evaluation datasets and production monitoring for retrieval quality, answer quality, latency, and failure modes.
I create a representative evaluation set that includes common workflows, difficult edge cases, ambiguous inputs, adversarial cases, and known failure modes. I combine deterministic checks, retrieval metrics, structured-output validation, model- or human-assisted qualitative review, and task-specific business metrics. Before release, I compare candidate changes against a baseline; after release, I monitor quality, drift, user behavior, error rates, and sampled outputs with a clear rollback path.
I would constrain the agent through explicit tool contracts, typed inputs and outputs, scoped permissions, idempotent actions where possible, timeouts, retry policies, and limits on steps and cost. High-impact actions should require confirmation or deterministic policy checks, while every tool invocation should be logged with trace context. I would also test tool failures, malformed model outputs, prompt injection attempts, and partial-completion scenarios before launch.
I would frame the choice in user and business terms: which tasks genuinely need the additional reasoning quality, what error reduction it delivers, and what it costs in latency and operating spend. I would propose measurable thresholds and an experiment, such as routing complex requests to the stronger model while using the faster model for routine classification or extraction. This gives Product a clear decision based on quality, responsiveness, and cost rather than model preference.
I first clarify the user problem, target workflow, harm scenarios, expected value, and measurable definition of success with Product and Design. I validate feasibility using a narrow prototype and representative data, then turn findings into an incremental architecture, evaluation plan, and launch criteria. I would ship a controlled first version, instrument it thoroughly, collect user feedback, and iterate based on observed quality and adoption.
About Pleo
Messy spend management is tricky business. And tedious processes are a lose-lose situation for all involved, not just finance. At Pleo, we’re changing that. We build spend solutions that make managing money seamless, empowering, and surprisingly effective for finance teams and employees alike – with a vision to help all businesses ‘go beyond’.
The word ‘Pleo’ actually means ‘more than you’d expect’, and living by that mantra has been the secret to our success over the last 10 years.
Now, we’re at a pivotal moment in our journey; every move we make has a direct impact on our 40,000+ customers, our business, and our collective success. We need people who take pride in uncovering customer needs, who turn complex problems into simple solutions, challenge the way things are done (respectfully), and always aim high. With great ambitions driving us forward, we can’t say we’ve got this whole thing figured out. And frankly, that’s half the fun! What we can say is that we’re a driven, progressive, and, importantly, a kind bunch of 850+ people from over 100 nationalities, all committed to delivering the future of business spending, together.
About the role
This is an exciting opportunity to help shape how Pleo builds AI-powered product features, working alongside software engineers, data engineers and data scientists to take ideas from prototype to production. Pleo has over 40,000 customers and a decade of unique spend data — an incredible foundation to build on. Your mission will be to harness this data to create real product value.
Please note: applications will be open until 1st September 2026 at 09.00 CEST. We will not review any application before the window is closed so please take all the time you need to submit a great application!
Who you’ll be working with and reporting to
You’ll be reporting to the Senior Manager for Data & AI Products and will be one of the first hires in the team but, rest assured, you will not be working solo! You will work very closely with other Engineers and Data Scientists, each bringing distinctive skills while you bring applied AI engineering and data context. Together, the team covers the full chain from data to shipped product.
Your core focus is on building and shipping AI-powered features, with a strong collaborative element across Product Engineering, AI Platform, and Data & ML Platform teams.
What you’ll be doing
As a Senior Applied AI Engineer you will:
Build and ship AI-powered product features for Pleo’s customers, owning outcomes end-to-end for a scoped area.
Design and build end-to-end RAG systems for specific product use cases: chunking strategies, embedding model selection, retrieval optimisation, and quality evaluation.
Own the full AI-product lifecycle: prompt design, context and state management, agentic loops, output parsing, edge case handling, and safe production deployment.
Build evaluation datasets and automated eval pipelines that give the team confidence in AI feature quality before and after changes.
Instrument AI features for production observability: logging, drift detection, quality monitoring, and alerting.
Collaborate with Product and Design to scope AI features from first principles; you are a co-author of what gets built, not just an implementer of specs.
Partner with our GenAI Platform team to identify infrastructure needs and champion adoption of platform tooling.
Support other engineers through reviews, pairing, and pragmatic technical leadership on the projects you lead.
What you bring
You will thrive in this role if you have:
Proven experience shipping customer-facing AI features using LLMs; beyond prototype stage and into real production systems with real users.
Deep practical experience with RAG system design and the full retrieval pipeline: embedding models, vector databases, chunking, hybrid search, re-ranking.
Experience building and operating evaluation frameworks for LLM-based systems; you have a systematic approach to measuring quality.
Strong Python engineering; your code is production-grade, tested, and maintainable.
Experience building APIs and data retrieval pipelines that feed context into AI systems.
Enough data intuition to reason about data quality, schema, and retrieval architecture without needing a dedicated data engineer beside you at all times.
Experience with agentic system patterns: tool use, multi-step orchestration, error handling in LLM workflows.
Familiarity with public cloud providers (AWS, GCP)
We operate a polyglot platform, with components written in Kotlin and Python. While this role is Python-focused, you will need from time to time to contribute in a Kotlin stack, so being open to that is important.
Why is this role a good fit for you
This role is a good fit for you if:
You have strong product instincts and build with the user in mind. You understand that a model is only as good as the problem it solves.
You have moved past prototyping and have a deep understanding of the realities of LLMOps, data retrieval, prompt and context engineering, as well as model evaluation in production.
You don’t just call APIs, you understand the data feeding the AI system and can reason about data quality, architecture, and retrieval without needing a dedicated data engineer beside you at all times.
This role is not a good fit for you if:
You are primarily interested in model research or algorithm development. This role is about shipping working AI features to production.
You need fully defined specs and a groomed backlog to operate effectively. We expect you to discover and define the problem alongside Product and Design.
You cannot explain AI trade-offs clearly to non-technical stakeholders. You will regularly work with Product Managers, Designers, and business leaders who need to understand what is and isn’t possible.
How you’ll develop in this role
In your first 6 months at Pleo, you’ll:
Familiarise yourself with our codebase, tooling and roadmap.
Partner with our Principal Engineer to define and own our approach to AI feature development.
Contribute to shaping the roadmap for tooling and features developed by our GenAI Core team.
Collaborate with Product and Data teams to ship a first feature to production
We’re committed to helping you develop your career, whether that means taking on bigger projects, stepping into leadership, or acquiring new skills!
The location
Please note: We can hire on a remote, hybrid or in-person set-up in any of the locations listed on the advert but you will need to be physically based in the country of your choice with a valid right to work.
We are unable to offer visa sponsorship for this role.
Show me the benefits!
Your own Pleo card (no more out-of-pocket spending!)
Lunch is on us for your work days – enjoy catered meals or receive a lunch allowance based on your local office
Comprehensive private healthcare – depending on your location, coverage options include Vitality, Alan or Médis
We offer 25 days of holiday + your public holidays
For our Team, we offer both hybrid and fully remote working options
We use MyndUp to give our employees access to free mental health and well-being support with great success so far
Paid parental leave – we want to make sure that we’re supportive of families and help you feel that you don’t have to compromise your family due to work
The interview process
We want to ensure you are set-up for success and understand what will be expected of you. If your application is successful, our interview process is as follows:
Intro call: A 30-minute chat with our Talent Partner to discuss the role and your background.
Technical screening: A 15 to 30 minute call with our Talent Partner to check your knowledge of key technical topics
System design interview: A 75-minute practical session with our engineers focusing on scoping and designing an AI feature
Live coding interview: A 75-minute practical session with our engineers focusing on implementing your solution.
Hiring Manager interview: A 60-minute conversation to deep dive into your knowledge and experience.
Final interview: A leadership interview focusing on your behavioural, communication and collaboration skills.
Transparency is important to us so we also wanted to share some insights about what we’re looking for in applications to ensure you can set yourself up for success!
CV writing and content: we receive a lot of CVs, and many of them are AI-generated. We love seeing people leverage AI—it’s a big focus for us internally too—but without human intervention, these CVs can sometimes become generic and fail to show a candidate in the best light. What we’re really looking for is the specific details of real impact that only you know from your previous experience. A top tip from us is to use the “Achieved X, as measured by Y, by doing Z” formula (credit: Laszlo Bock, ~2014) to give a really clear picture of what you’ve worked on. A final note: including links to your previous companies’ websites is a huge help and allows us to truly understand your background!
Application care: every single application we receive is reviewed by a human (yes, hundreds of them) because we believe that candidates’ efforts should be matched by an equal level of human care. This means that we expect a similar level of attention put into your application. Read and answer the application questions carefully, they make a huge difference in our decision-making process.
Profile to role fit: this is neither a research-driven nor a traditional data or software engineering role. We’re looking for someone with proven experience shipping customer-facing AI products who can also confidently operate at Senior level in a SaaS organisation. This means we’ll be looking for evidence that you can own fairly complex technical features from design to implementation.
About your application
English first. Since it’s our company language, please submit your application in English. You’ll be using it a lot if you join us.
A fair look for everyone. Our talent team reads every single application to ensure the process is fair. To keep things running smoothly, we only accept applications through our system—our support team can’t pass on calls or emails.
Diversity drives us. We can only reach our goals if our team reflects the world around us. That starts with you hitting apply, even if you don’t tick every single box. We encourage people from all backgrounds and experiences to join us.
Interview at your best. We want you to feel comfortable throughout the process. If you have any accessibility requirements or need a specific format, email belonging@pleo.io. We’ll design a process that works for you.
Your data is safe. When you apply, we process your personal data as a data processor. For more information on how Pleo processes personal data, read our Privacy Policy here.
Applying for multiple roles? Nothing is stopping you, and we assess every role independently. However, we do look for alignment, so make sure you can explain why your interest and experience are right for each specific role.
Reapplying. If you’re applying for the same role again, please wait six months from your last decision before hitting submit.
Annual salary information is not provided for this position. Explore salary ranges for similar roles in our Salary Directory ›
Upload your resume now
- To unlock remote work opportunities and be discovered by global employers.
This job listing has been manually reviewed by the Jobicy Trust & Safety Team for compliance with our posting guidelines, including verification of the company's legitimacy, accuracy of job details, clarity of remote work policy, and absence of misleading or fraudulent content.
