Staff Engineer: Agentic Reasoning & Core Capabilities

Arena Physica; New York (NY); Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Health insurance (Aetna)

401(k) Retirement Plan

Unlimited PTO

Lunch via Sharebite

Relocation support

Job summary

Arena Physica in NYC is seeking a senior software/ML systems engineer to help Atlas reason, plan, and operate across hardware design lifecycles. You will turn expert electrical workflows into agentic capabilities and drive end-to-end workflows in real engineering environments.

The role emphasizes building reliable tool-using agents, improving model behavior in production, and collaborating with researchers, product teams, and field engineers.

Qualifications

• 5+ years of software engineering, research engineering, or ML systems experience at a venture-backed startup, top technology company, frontier AI lab, or similarly demanding environment.
• Demonstrated experience building LLM-powered systems, agentic workflows, model tooling, evals, retrieval systems, or production AI products.
• Strong software engineering fundamentals and the ability to move fluidly between prototypes, infrastructure, product surfaces, and production services.
• Excellent judgment around model behavior, user experience, reliability, observability, and the failure modes of AI systems.
• Comfort operating in ambiguous problem spaces where the right abstraction is discovered through iteration with users, data, and model behavior.
• Ability to collaborate deeply with researchers, product engineers, electrical engineers, and customer-facing teams.
• Strong communication skills and a track record of driving complex technical projects from idea to shipped capability.
• Self-directed ownership mindset and excitement for building core systems at the frontier of AI and hardware engineering.
• [Preferred] Experience with tool-using agents, coding agents, computer-use agents, multi-agent systems, RL/post-training, synthetic data generation, harness engineering, or automated evals.
• [Preferred] Experience building AI systems for technical users such as engineers, researchers, analysts, developers, or operators.
• [Preferred] Familiarity with hardware engineering, electromagnetic simulation, RF systems, signal integrity, power integrity, EDA tools, lab instrumentation, or manufacturing workflows.

Responsibilities

• Build Atlas's reasoning core - Design and implement agentic planning, tool use, memory, retrieval, self-checking, and workflow execution capabilities for hardware verification, debugging, and optimization.
• Turn expert workflows into an agentic harness - Translate EM, RF, signal integrity, power integrity, lab validation, simulation, and field-repair processes into reusable agent capabilities that work across real customer environments.
• Improve model behavior in production - Develop systems that make Atlas more reliable at following long-horizon tasks, asking the right clarifying questions, grounding answers in evidence, and recovering when tools or assumptions fail.
• Create domain-specific evaluations - Build evals, graders, traces, and diagnostics that measure agent performance on hardware engineering tasks rather than generic chatbot benchmarks.
• Connect models to tools and data - Integrate LLMs, electromagnetic foundation models, retrieval systems, solvers, lab data, design artifacts, and engineering systems of record into coherent end-to-end workflows.
• Ship with customers in the loop - Partner with Product, Go-To-Market, and forward-deployed engineering teams to learn from real deployments and rapidly convert feedback into stronger reasoning systems and core capabilities.
• Travel domestically and internationally (10-20% of your time).
• Work in person at Arena's NYC HQ when not traveling.

Skills

LLM-powered systems

Agentic workflows

Tool-using agents

Production AI systems

Software engineering fundamentals

Cross-functional collaboration

Tools

LLM tooling

Retrieval systems

Solvers

Lab data systems

Job description

Arena Physica in NYC is seeking a senior software/ML systems engineer to help Atlas reason, plan, and operate across hardware design lifecycles. You will turn expert electrical workflows into agentic capabilities and drive end-to-end workflows in real engineering environments.

The role emphasizes building reliable tool-using agents, improving model behavior in production, and collaborating with researchers, product teams, and field engineers.