Staff Software Engineer - AI

About Element34

For years, Element34 has run large-scale test execution for enterprises that can't send their testing to a public cloud. Banks, insurers, healthcare companies and other regulated organizations use our platform, SBOX, to run thousands of Selenium, Playwright and Appium sessions in parallel on their own infrastructure.

Now we're building an AI-native platform for the whole end-to-end test lifecycle: authoring, execution, failure analysis, triage and self-healing maintenance, with LLMs and agents at the core of each stage rather than added on afterward.

We're a small, senior engineering team and we plan to stay that way. We have no layers of management, no handoffs between teams and no separate infra group. Each engineer owns a problem from design through production.

The Role

This is a hands-on Staff Engineer role with no direct reports. Your influence comes from the systems you design, the code you write and how you raise the engineering bar for the people around you.

About 90% of your time will be backend work in Java or Go: distributed systems, concurrency, data pipelines, AI and agent infrastructure, and performance work. The other 10% is frontend work in React.

This is not CRUD or API plumbing. You'll work on orchestration, scheduling, high-throughput ingestion, retrieval quality, agent reliability and performance problems that don't have a ready-made answer.

What You'll Do

• Build the AI core of the platform. Design and ship LLM-powered features and autonomous agents that author tests, diagnose failures, find root causes and repair broken tests.
• Build RAG and retrieval systems over test artifacts, logs, traces, DOM snapshots and code. You'll own chunking, embeddings, hybrid search, ranking and context construction, and measure retrieval quality rigorously.
• Make AI production-grade. Build evaluation harnesses, guardrails, observability, cost and latency controls, and fallbacks so agents behave predictably in enterprise environments, including self-hosted and air-gapped ones.
• Design and build core platform systems in Java or Go, including session orchestration, scheduling, resource management and high-concurrency services behind thousands of parallel browser and device sessions.
• Own the data layer across PostgreSQL (transactional), ClickHouse (analytics and high-volume telemetry), OpenSearch (logs, search and vector retrieval) and Redis (caching, coordination, queues).
• Own delivery across the whole lifecycle. We have no separate infra team. You'll containerize with Docker, deploy and run on Kubernetes, automate with Ansible, and make sure the software runs reliably in customer environments you don't control.
• Solve hard problems end to end. Find the bottleneck, profile it, design the fix, ship it and watch it in production.
• Ship features in React when the work needs it, with no frontend team to hand off to.
• Set technical direction through design docs, architecture reviews and prototypes. Make the hard trade-off calls and explain them clearly.
• Make the team AI-native. Use AI coding tools every day and help shape how the team builds with them.

What We're Looking For

• Strong professional software engineering experience, with a strong record of building and owning complex backend systems.
• Deep expertise in Java or Go, including concurrency, memory management, performance tuning and idiomatic modern code. You write code others learn from.
• Mandatory: hands-on experience building with LLMs, including:
• RAG systems in production or close to it: embeddings, vector or hybrid search, chunking strategies, reranking and retrieval evaluation.
• AI agents: tool and function calling, multi-step planning, memory and state, and handling failures and non-determinism. Experience with MCP or agent frameworks is a plus. Knowing when to skip a framework counts too.
• Prompt and context engineering, plus evaluation. You know how to measure whether an LLM feature actually works.
• Daily, effective use of AI coding tools (Claude Code, Cursor, Copilot or similar).
• Strong computer science fundamentals and problem-solving ability. You can reason about correctness, performance and distributed-systems trade-offs from first principles.
• Production experience with data stores. Solid PostgreSQL skills plus hands-on experience with at least one of ClickHouse, OpenSearch/Elasticsearch or Redis at scale.
• Hands-on with Docker, Kubernetes and Ansible. You've deployed, debugged and operated your own services.
• Working proficiency in React and modern frontend development.
• An owner's mindset. You do well with autonomy and ambiguity, and you'd rather fix a problem than escalate it.
• Clear written communication. You're comfortable working remotely and asynchronously, where a good design doc replaces a dozen meetings.

Nice to Have

• Experience with browser automation and test infrastructure: Selenium/WebDriver, Playwright, Appium, Selenium Grid, or device and emulator farms.
• Deeper Kubernetes knowledge: operators, custom controllers, autoscaling, or multi-tenant design.
• Experience with high-throughput ingestion and analytics, such as event pipelines, time-series data, or ClickHouse schema and materialized view design.
• Experience running self-hosted or open-weight LLMs (vLLM, Ollama, TGI) for air-gapped or data-sensitive deployments.
• Experience with fine-tuning, small-model distillation, or multimodal models (screenshots, DOM, visual diffing).
• Background building on-premise enterprise software for locked-down or regulated customer environments.
• Familiarity with security and compliance expectations in financial services or healthcare.

Why Element34

• Build an AI-native product, not add AI to an old one. You'll shape how end-to-end testing works with agents at the center.
• Real ownership. Small team, large scope, and your work reaches production at major enterprises.
• Hard problems. Agent reliability, retrieval at scale, distributed orchestration and performance.
• Fully remote, with a team that values deep work and written communication.
• No bureaucracy. Nothing stands between you and the decisions, the customers or the codebase.