Staff Engineer
TypeScript, Node.js, PostgreSQL, LLM Systems
Why join Levellr
Levellr is the enterprise community intelligence and management solution for Discord and Reddit, used by some of the world’s largest gaming companies, including Scopely, Krafton and Epic, as well as brands such as Google, YouTube and SoundCloud, to help them grow, manage and monetise their communities.
Discord and Reddit communities can generate millions of messages, but for the teams running them, it is hard to turn all of that activity into useful insight. Levellr helps them see what matters, understand their members, spot trends, improve engagement and make better commercial decisions from their community data.
The product is now moving into a more technically demanding phase. We are building AI-powered systems into the core of Levellr, including agents, evaluation pipelines, anomaly detection, cost infrastructure and LLM-powered workflows. These systems need to make sense of large, messy, fast-moving community data, and they need to work properly in production.
What you’ll work on
A big part of the role is leading the architecture and delivery of foundational systems across Levellr’s AI and data platform.
That includes production agent systems, evaluation pipelines, anomaly detection, cost infrastructure, data models, orchestration patterns and internal frameworks that help the rest of the team build faster and with more confidence.
The data side really matters - Levellr processes millions of Discord and Reddit messages, so we need someone who understands what it takes to design, tune and evolve relational systems at scale. PostgreSQL is a big part of that. Indexing, partitioning, query performance, schema design, migrations and data modelling are central to the role, not just useful extras.
The AI side needs to be practical too - We need someone who has seen what happens when LLM systems meet real users, real data, real cost and real failure modes. You will help shape how Levellr thinks about agents, model behaviour, evaluation, quality, observability, cost control and recovery patterns.
You will work closely with product, design, customer success and leadership. Some problems will be clearly scoped. Many will not be. A lot of the value in this role comes from taking a vague problem space, working out what matters, and turning it into something useful that ships.
Wider team impact - The right person will become a technical reference point for other engineers. Not by creating lots of processes or sitting above the work, but by building patterns, writing clear PRs, sharing good Looms, making sensible architectural calls, and helping the team move faster without getting loose.
How we ship
Levellr has a strong bias towards shipping and learning from production.
We do not wait until everything feels perfect. We ship the next sensible version, get it into real usage, then improve based on what we see.
Some of our most important systems, including Levellr AI, agent quality evaluation and anomaly detection, have gone from blank page to shipped foundations in weeks, not quarters.
That pace only works if the engineering judgement is strong.
Good engineering here means knowing when to keep something simple, when to invest properly in foundations, when to refactor, when to ship and come back, and when to admit the first approach was wrong.
The level we are looking for
Strong candidates will have deep experience with relational data at real scale. That means production systems where volume, query performance, indexing, partitioning, schema design or database architecture genuinely mattered.
You should have built or led technical work that other engineers, teams or products depended on. This could be platform work, data infrastructure, pipeline orchestration, evaluation systems, cost infrastructure, architecture refactors or similar foundational work where the impact compounds over time.
Production AI or LLM experience is important. We are not looking for light API integrations, side projects or prompt experiments. We need someone who has worked on AI-powered systems where quality, evaluation, cost, observability and edge cases all became real engineering problems.
Staff-level autonomy is a key part of the bar. You should be able to spot the issue, frame the trade-offs, propose a path, bring people with you and get the system shipped without needing lots of senior people to create the structure for you.
Systems thinking matters. The work cuts across product, data, infrastructure, customer value, reliability, cost and team velocity. We need someone who can see the wider shape of the system, not just the immediate ticket.
What good looks like in you
You have around 10+ years of professional software engineering experience, or enough depth to show you are already operating at that level.
TypeScript and Node.js are very familiar to you, and you are comfortable working across the stack when needed.
PostgreSQL is a real strength. You have worked with large relational datasets and can talk clearly about indexing, partitioning, query optimisation, schema design and migration trade-offs from hands-on experience.
You have shipped production AI or LLM-powered systems. Ideally that includes agents, evaluation frameworks, RAG, prompt engineering at scale, cost optimisation, model behaviour or AI product infrastructure.
Essential: You have proven, at-scale experience with customer-facing LLMs, specifically focused on testing and iterating based on real-world end-user interactions.
You are still very hands-on. You can lead the thinking and write the code.
Ambiguity does not throw you. You can take a loose product or customer problem, work out what needs to be true technically, and move it forward without waiting for a perfect brief.
You ship quickly, then use real signal to improve the system. Your instinct is not to disappear for weeks and come back with a big reveal. You prefer tight loops, visible progress and practical learning.
You make other engineers better through how you work. Your PRs, notes, Looms and technical decisions give people useful reference points, not extra noise.
You want to work as part of a fully remote, European-timezone team, with meaningful overlap with the UK working day.
This role probably is not right if
Most of your experience has been building standard SaaS product features without much data complexity.
Your AI experience is mostly demos, side projects, experiments or basic API integrations.
You have not worked with relational data at enough scale for indexing, partitioning, schema design and query performance to become serious problems.
You need a lot of structure before you can make progress.
You prefer long planning cycles before anything reaches users.
You want a pure architecture role where other people do most of the implementation.
You are not comfortable making technical trade-offs quickly.
You have not yet owned technical decisions where the wrong call could affect customers, cost, reliability or team velocity.
The tools we use
TypeScript, Node.js, NestJS and SvelteKit across the product.
PostgreSQL, ClickHouse, BigQuery, Redis and PubSub across data, analytics and messaging.
Google Cloud, Kubernetes, GitHub Actions and Cloudflare for infrastructure and delivery.
OpenAI, Anthropic, Vertex AI and other LLM platforms across our AI systems.
Nice to have
Experience with Vertex AI, OpenAI, Anthropic or similar LLM platforms at scale.
Experience with Temporal or similar durable workflow systems.
Experience with ClickHouse, BigQuery or other analytical databases.
Experience building agent architectures, evaluation systems, anomaly detection or LLM cost infrastructure.
Previous startup or growth-stage experience where you have owned technical direction and speed.
Strong technical writing, talks, open-source work or internal engineering frameworks that other people have actually used.
How we work
Levellr is remote-first, with flexible working hours and a sensible expectation of overlap with the UK working day.
Engineering works in small, focused squads, closely with product and design.
We use Linear for planning, Slack and Discord for day-to-day communication, and Loom for async demos and walkthroughs.
The team suits people who like ownership, clear thinking and practical progress. There is plenty of ambiguity, but not much appetite for unnecessary process.
People are trusted to do the right thing, communicate clearly and keep momentum without needing lots of oversight.
Benefits
Highly competitive basic salary - full details shared at the initial interview stage
25 days annual leave plus UK public holidays.
Remote or hybrid working, with flexible working hours.
Co-working space access, including Soho Works locations across London and the US.
Stock option plan.
Home office budget once you pass probation.
Mental health support through Spill.
Twice-yearly company meet-ups.
Interview process
Initial screening call with our Talent Partner (Matt) to talk through your background, the role and whether there is a strong fit both ways.
Introductory conversation with our Engineering Manager (Miro) to give more context on the product, technical direction and how the team works.
Technical pairing or technical deep dive focused on how you think, make trade-offs, and ship working systems. This stage includes a complex, AI-focused technical task designed to assess your iterative shipping mindset, collaboration with Product, prioritization skills, and tolerance for ambiguity.
Staff-level technical interview with Ben, our CTO, and the engineering leadership team. This will go deeper into relational data at scale, production AI systems, architecture, decision-making and examples of where you have led foundational technical work.
Final conversation with senior leadership to cover ways of working & culture fit
We aim to complete the process within two weeks and provide clear feedback after each stage.
Originally posted on