Staff SW Engineer

Description

The company builds specialized, efficient AI models and is at the seed stage, having raised $25M led by a group of investors that includes Microsoft, Khosla Ventures, Insight Partners, GitHub CEO Thomas Dohmke, and Docker CEO Scott Johnston, among others.

The team includes alumni of Google Research, Apple, Stanford, and Cambridge.

The company's GLiNER family of open source models has been downloaded over 50 million times and is used by NVIDIA, Meta, and Airbnb, among others.

This is a Staff Software Engineer role.

Responsibilities

• Help set the technical direction for the platform and shape architecture decisions so the system can grow without slowing the team down.
• Design and build the agent harness, covering the runtime loop, tool calling, context management, and error recovery, along with the eval harnesses that measure agent reliability.
• Build and scale AI training and inference pipelines, improving latency, throughput, and cost as usage increases.
• Own deployment of the platform into customer-managed AWS VPCs with Kubernetes and Terraform, including install, upgrade, and observability paths for environments the company does not control.
• Build secure sandboxed execution environments for agentic workloads on Modal Sandboxes.
• Ship full-stack features from start to finish, taking on the hardest and least defined problems on the team.
• Work closely with AI researchers to move new models from research into production, including deployment, evaluation, and monitoring.
• Raise the engineering bar through code reviews, mentoring, and setting practices others on the team will follow, including how the team works with AI coding agents.
• Talk with customers and turn their feedback into product decisions quickly.

Requirements

• Required: 8+ years of hands-on software engineering experience, including time as a tech lead or leading architecture across a team or product.
• Required: Strong architecture skills and a record of designing and scaling production systems under real load.
• Required: Production experience with agentic systems, AI training pipelines, or inference pipelines, ideally more than one.
• Required: Hands-on experience deploying software on AWS with Kubernetes and Terraform, ideally into customer-managed or multi-tenant environments.
• Required: Experience building and running sandboxed execution environments for agent workloads, ideally with Modal Sandboxes.
• Required: Extensive experience working with AI coding agents, and the ability to guide them toward clean, production-quality code. You can scope tasks, review output critically, and catch agent mistakes.
• Required: Comfort with ambiguity. You make sound decisions without a playbook and know when to move fast and when to build for the long term.
• Optional: Experience building agent harnesses or frameworks used in production, including familiarity with MCP or similar tool protocols.
• Optional: Experience with LangSmith or similar tooling for tracing, evaluating, and monitoring agents.
• Optional: Hands-on experience with inference optimization, such as batching, quantization, GPU utilization, or serving frameworks like vLLM, SGLang, or Triton.
• Optional: Deep experience with high-concurrency backend systems or low-latency API design.
• Optional: Startup experience, ideally as an early engineer helping a team grow.