Software Engineer (AI Infrastructure)

Software Engineer - AI Infrastructure

Conditions

Posted Date May 19, 2026 Employment Type Full-time Experience Level Entry level Location San Francisco Bay Area Annual Salary 220,000 USD Category ** Programming ** Company **Aurora **

Software Engineer - AI Infrastructure

Software Engineer — AI Evaluation Infrastructure

San Francisco, CA · On-site · Full-time$180K–$220K base + $230K on-target bonus + uncapped upside

The company

The company builds infrastructure for frontier AI evaluation , novel dataset creation, and experimentation. Its systems support agentic workflows and hard-reasoning use cases, where the product only works if the underlying data, experiment execution, and annotation flows stay reliable.

It works with all five leading AI labs and is becoming the go-to partner for YC companies that need serious AI data infrastructure.

Founded in 2024 , the team is still small—about 20 people —with $34M raised and a background that includes quantitative trading, Google, Meta, Morgan Stanley, Silver Lake, Jane Street, Goldman Sachs, and Bloomberg. Although the business is B2B, the engineering problems are closer to platform infrastructure than conventional SaaS feature work.

The role

This is an end-to-end engineering role for someone who wants broad ownership across product surfaces, backend systems, and infrastructure.

One week you might turn a research idea into a new RL environment. The next, you might deploy distributed experiments on Kubernetes. The week after that, you might improve the reliability of a Next.js dashboard or build a Kafka pipeline for annotator analytics.

You will work directly across research, product, and design, so the job requires translating ambiguous ideas into concrete interfaces and system boundaries.

The expectation is not to stay in one layer of the stack. You will own the systems that turn research ideas into production workflows, customer-facing tooling, and repeatable operations.

The technical problem

AI evaluation is an orchestration problem disguised as a product problem.

You have to coordinate compute, data creation, human annotation, experiment tracking, and dashboarding while keeping the system observable and correct under real usage.

The system has to be accurate enough for labs to trust, fast enough for operators to use daily, and flexible enough to absorb new experiment types without constant rework.

The hard parts are reproducibility, schema discipline, retry logic, throughput, data quality, and making sure the team can move quickly without breaking trust in the results.

What you'll own

• RL environments and experiment systems: prototype and ship environments from research ideas, then make them reliable enough for repeatable use.
• Human-in-the-loop workflows: build the systems that manage annotations, reviewer flows, and the data paths that feed evaluation. Searching for Development & Programming roles that provide visa sponsorship? Connect with international employers through Development & Programming Jobs with Visa Sponsorship opportunities actively seeking talented professionals.
• Distributed infrastructure: deploy and operate workloads on Kubernetes with a focus on reliability, reproducibility, and debuggability.
• Data pipelines: build Kafka-based pipelines and supporting services for annotator analytics and other high-volume event flows.
• Product surfaces: keep the Next.js dashboards and related tooling reliable, fast, and useful for the people operating the system.
• Core APIs and services: design the backend interfaces that connect research, product, and operations without creating one-off glue code.
• Engineering quality: write maintainable code, participate in design reviews, and document the decisions other engineers will have to live with.

Who this is for

You are likely a strong fit if you have:

• Owned production systems end to end, not just features inside a larger platform.
• Built systems that involve distributed execution, queues, retries, observability, and backfills.
• Strong judgment about data models, schema evolution, and where to normalize versus denormalize.
• Experience shipping across both application code and infrastructure.
• Comfort moving between TypeScript/JavaScript, Node.js, Python, and Kubernetes without treating stack changes as a blocker.
• The ability to work from ambiguous requirements and turn them into concrete system boundaries.
• Good instincts for reliability, correctness, and maintainability when the right answer is not the fastest one.
• The communication skills to work with research, product, and design without losing technical rigor.
• A preference for broad ownership over narrow ticket execution.
• Experience building internal platforms or data infrastructure where reliability and observability mattered at scale.

Tech stack

• Frontend: Next.js, React, TypeScript / JavaScript
• Backend: Node.js, Python
• Infrastructure: Kubernetes, cloud platforms on AWS and GCP
• Data and systems: Kafka, Redis, Elasticsearch

The team works across the stack because the product spans user-facing dashboards, backend services, experiment orchestration, and data infrastructure.

Why now

The company is at the point where research prototypes are no longer enough on their own.

The next constraint is building infrastructure that can support more customers, more experiments, and more annotation throughput without adding operational drag.

This is the stage where architecture decisions compound: the data model, queueing strategy, observability, and service boundaries will shape how quickly the team can ship for the next several years. Explore our comprehensive directory of visa sponsorship jobs from employers worldwide who are ready to sponsor talented international professionals.

This role is not for you if

• You only want isolated feature work with clearly defined tickets.
• You prefer staying in a single layer of the stack.
• You do not want to work on infrastructure that affects data quality or experiment correctness.
• You need extensive process before you can make a decision.
• You are not comfortable being on-site in San Francisco.

Compensation and logistics

• Base salary: $180K–$220K
• On-target bonus: $230K
• Upside: uncapped; the bonus structure includes 4% of profits tied to the value of total contracts in your vertical
• Commercial context: average contract value per vertical is $8M, with average margins around 70%
• Equity: competitive
• Visa support: available
• Base flexibility: available for unusually strong backgrounds, with sign-on bonus support possible
• Location: San Francisco, CA
• Work model: on-site 5 days a week in FiDi (M-F), with a half-day or remote Sunday
• Employment: full-time

Interview process

• Initial screen — 30 minutes: background, scope, and fit.
• Engineer interview: technical depth and systems thinking.
• Take-home and review: about 6 hours of real work.
• Work trial: 2 days in the SF office.
• References: if necessary.

About Aurora

Aurora helps exceptional engineers find the right role at some of the most ambitious startups worldwide.

We work with teams that value high ownership, strong technical standards, and clear scope.

Показать контакты