Software Engineer, AI Systems
Making agentic analytics reliable by design.
At Veezoo, we care deeply about making agentic analytics reliable: answers should be correct, traceable, understandable, and consistent enough to use in real business decisions.
Getting there involves much more than prompting. Our product brings together language and compiler technology, agentic systems, harness engineering, evaluation, data modeling, and product design. As an engineer, you'll work across those boundaries.
This is a broad engineering role. In one week, you might work on how we translate a user's question into a structured query; in another, you might improve an agent's planning or evaluation loop, or refine how an analytical result is presented to the user.
We're a small team, so the problems you work on are driven by what the product needs most. You'll have a lot of ownership and work closely with the rest of engineering and product.
Veezoo is an ETH Zurich spin-off based in Zurich. Our customers include companies such as AXA, BMW, Bayer, and Breitling, which means we work with real-world data models, complex business logic, governance requirements, and high expectations for accuracy.
We're building a business intelligence product that companies can trust as part of their everyday decision-making. Reliability isn't a layer we add at the end; it shapes how we design and engineer the system from the ground up.
Why this role is different
A language designed for reliability
VQL is a domain-specific language with a parser, type system, optimizer, and compiler targeting every warehouse we support. It gives the model a precise language for expressing analytical queries, with clear constraints on what can and cannot be represented.
That means reliability is something we can engineer into the system: through language design, static checks, compilation, and evaluation, not just model behaviour.
A knowledge graph as the source of truth
Business meaning is modelled once in VKL, our modelling language, and versioned in Git with branches, diffs, and automated tests. Data teams work in Veezoo Studio, the IDE we build for them.
Designing an ontology that is expressive for analysts, maintainable by data teams, and understandable by a language model is a hard problem. You would own parts of that design.
AI is part of how we build
We use coding agents heavily in our own engineering workflow, and we expect every engineer to know how to work with them well: where they accelerate the work, where they fail, and how to review their output critically.
The same applies to the product. We run automated evaluations over model and prompt changes, route requests across model tiers to manage cost and latency, and use AI throughout our internal tooling and content workflows.
Engineers shape what gets built
There is no handoff from a product manager with a finished specification. Engineers talk to customers, look at real usage, form opinions about the product, and ship.
The interface a business user sees matters just as much as the query planner behind it. We care about both technical depth and product craft, and expect engineers to take responsibility for the whole thing.
The stack
- Backend
- Scala on the JVM, PostgreSQL, a query compiler we wrote ourselves
- Frontend
- TypeScript, React
- AI
- Frontier and open-weight LLMs, agentic orchestration, an in-house evaluation harness
- Warehouses
- Snowflake, BigQuery, Databricks, Redshift, SAP HANA, Trino, and a dozen more
- Deployment
- Fully managed cloud, or single-tenant in a customer AWS, Azure, or GCP region
What you bring
- Several years of production software engineering experience, including substantial work you owned end to end
- Strong computer science fundamentals. Data structures, algorithms, system design, and abstraction matter more to us than any particular framework
- Willingness to work across the stack. You do not need to know Scala, but you should enjoy strongly typed code and be able to learn it quickly
- A real working practice with AI coding tools: you know when agents help, when they fail, and how to review their work
- Product judgment and a willingness to form and defend opinions
- Fluent English; German is useful but not required
- A valid Swiss work permit, or eligibility for one as an EU/EFTA citizen
Bonus skills
- Compilers, parsers, type systems, query engines, or database internals
- LLM systems in production: evals, latency budgets, cost control, non-determinism
- Knowledge graphs, ontologies, or semantic modelling
- Data and BI experience: warehouses, dbt, semantic layers, or years spent living inside a BI tool as a user
- You've built and shipped something of your own that people actually use
How we work
- Zurich office, hybrid. We spend a good part of the week working together in person
- Small team, no layers. You work with the founders every day
- We ship continuously. Your work reaches enterprise customers in days, not quarters
- Equity options, so the outcome is shared
This role may not be a fit if
- You prefer working from detailed specifications within a narrowly defined area
- You want to write every line of code yourself rather than use agents as part of your workflow
- You are looking for a fully remote role; this position is based in Zurich
- You prefer the structure and predictability of a large engineering organization; we are a small team and priorities can change quickly
Interested?
Write to join@veezoo.com. Tell us what you've built and what you'd want to work on here. A repository, side project, or a short note about something you hold a strong technical opinion on tells us more than a polished cover letter.
Apply for this role