Model Systems Engineer

  • Location: Remote
  • Department: Platform
  • Employment Type: Full-time

What’s in it for you?

Most companies building on AI right now wrap a single model in a thin layer and call it a product. You would own the layer underneath that. Paciva sends reasoning across many models, cloud and local, and decides which one takes which job at which level of effort. That routing is why the platform runs faster and costs less than the companies betting everything on one provider.

The first version works. It is one person deep. You would take it from a founder’s build to a system a company can stand on, and you would set how good it gets.

About the team and the opportunity

Pax is an executive assistant for individuals and for entire organizations. It directs message intelligence [Live], inbound screening [Live], relationship and trust work [Live], and voice [Live]. Every one of those sends reasoning through the layer you would own.

Paciva is token optimized by design. The platform sends the model only what it needs to reason, which means quality, speed, and cost all move when you change how routing works. Few engineering jobs give one person that much leverage over a product’s economics.

A day in the life

You open an overnight evaluation run and see one model class drifting on a task it used to handle well. You trace it, adjust routing, and rerun. Midday you sit with the founders on local model support, because a customer on regulated data needs work to stay on their own hardware. You spend the afternoon on the cost curve, because a change that saves fractions of a cent per call becomes real money at scale.

What you will work on

  • Own model routing across cloud and local, including quality tiers and level of effort
  • Build and maintain evaluation runs that catch drift before customers do
  • Tune context so the platform spends the fewest tokens that still produce the right answer
  • Add and retire models as the field moves, with a documented process, not guesswork
  • Set up failover so a provider outage never becomes a customer outage
  • Publish the quality metrics the rest of the company plans against
    Work with product on which capabilities need frontier reasoning and which do not

What we need from you

  • 5+ years building production systems, with at least one year on LLM infrastructure
  • Real experience with evaluation, not vibes, and the discipline to trust the numbers
  • Comfort running open models locally, including quantization and hardware limits
  • Strong instincts about cost and latency tradeoffs under load
  • Ability to explain a routing decision to a founder and to a customer
  • Bias toward measurement over opinion, and toward simple designs that hold up

Remote work, compensation, and benefits

This is a remote role. Paciva is based in Miami, Florida & Austin Texas and the team works across time zones. We value clear written communication, ownership, practical collaboration, and dependable follow-through. Compensation is based on experience, skills, location, market data, role scope, and internal equity. For sales roles, compensation may include base salary and variable incentive compensation. Final details will be shared during the hiring process. Most roles include occasional travel for team meetings, customer sessions, company events, or off-sites.

Equal opportunity statement

Paciva is an equal opportunity employer. We make employment decisions based on qualifications, merit, business needs, and role fit. We do not discriminate based on race, color, religion, sex, sexual orientation, gender identity or expression, national origin, age, disability, veteran status, genetic information, marital status, pregnancy, caregiver status, or any other status protected by applicable law. If you need a reasonable accommodation during the application or interview process, please let us know. We will work with you to support your participation.