Careers at Zentrik
8 senior roles
Agent-native brains
All roles

The Evals Brain

Make Zentrik better at knowing what it knows, where evidence came from, and whether agentic work is improving.

Read role detailsStart a conversation
Data systems
Innovative role

The Evals Brain

Data quality, knowledge systems, model behavior, and evaluation practice

A data engineer, AI evaluator, and product systems thinker shape.

Make Zentrik better at knowing what it knows, where evidence came from, and whether agentic work is improving.

Start a conversation

A useful first note shows how you think and what you have made.

What you would own

  • Data quality, source lineage, context modeling, evaluation sets, and feedback loops for product intelligence workflows.
  • Systems that measure model behavior, recommendation quality, retrieval quality, cost, drift, and customer-facing reliability.
  • Tools and review practices that make non-deterministic work more inspectable, comparable, and trustworthy.

What we would look for

  • You can reason across data modeling, product semantics, eval design, analytics, and AI workflow reliability.
  • You understand that evals are product infrastructure, not an afterthought.
  • You can build practical harnesses that help humans decide whether the system is getting better.

Questions you would help answer

  • How should Zentrik know that a product recommendation is actually improving?
  • What evidence and lineage should follow an insight from signal to shipped work?
  • How should evals, analytics, and human review reinforce each other?
What the work feels like

The product is the operating layer for better product decisions.

The role is not to add AI decoration to old workflows. The role is to preserve product intent from customer evidence to shipped work, while making every loop more inspectable, more reliable, and easier for a human team to trust.

Product intent stays human

AI can accelerate the work, but the reason to build still comes from evidence, taste, tradeoffs, and accountability.

Visible systems beat claims

We look for artifacts: GitHub repos, workbenches, eval notes, customer loops, market systems, prototypes, and operating docs.

The harness matters

Great AI-native work depends on context, evals, guardrails, review loops, and proof that the next cycle gets sharper.

Zentrik agent workflow graph showing product context moving into agent work.
How we think about fit

Range matters, but judgment matters more.

We are more interested in visible systems of work than in claims of AI fluency. Strong candidates can show how they create leverage without confusing output volume for customer value.

GitHub or public work artifacts

Agent harness or eval loop

Product or business judgment

Market or customer learning system

Infrastructure or MLOps system

Community or enablement motion

Personal workbench or knowledge system

Hiring process

Show us the system behind the work.

A strong note could include a GitHub repo, product teardown, prototype, GTM system, research synthesis, eval harness, personal workbench, shipped product, or a short memo about where you think Zentrik should go.

Contact Zentrik
Tell us which role fits your strongest proof.
  1. 1

    Send a proof artifact

    Share a repo, product teardown, prototype, GTM system, research synthesis, eval harness, customer system, or short memo.

  2. 2

    Work through the role shape

    We will talk about the domain, the level of ownership, the systems you have built, and where your judgment is strongest.

  3. 3

    Review a real operating problem

    The useful signal is how you structure ambiguity, design feedback loops, and decide what is worth doing next.

  4. 4

    Define the first scope

    If there is a fit, we will shape the initial domain clearly enough that you can start creating leverage quickly.