Software Engineer

Who you are

• 3+ years of software engineering experience building production distributed systems or cloud‑native applications
• BS/BA in Computer Science or related field, or equivalent practical experience
• Strong coding skills in at least one of: Python, Go, Java, Javascript, Typescript, C++, with a focus on reliability, performance, and tests
• Product‑minded: you prioritize customer impact, clear SLAs/SLOs, and pragmatic iteration
• Ownership‑driven with a positive, proactive attitude; comfortable leading projects or learning from battle‑tested engineers
• Experience operating services on Kubernetes and at least one major cloud (e.g., GCP, AWS, or Azure)
• Familiarity with event/streaming systems (e.g., Pub/Sub, Kafka), caching (e.g., Redis), and data stores for low‑latency paths
• Practical understanding of LLM/agents building blocks: tool/function calling, structured outputs, streaming, and model selection/routing
• Strong observability and debugging skills: tracing (e.g., OpenTelemetry), metrics, dashboards, and production forensics
• Background in one or more areas is a plus: policy/guardrails, multi‑tenant isolation, rate‑limiting, concurrency control, cost optimization

What the job involves

• The Agentic Runtime team builds the low‑latency, reliable, and secure foundation that powers Glean’s AI agents and assistant experiences at scale
• You’ll design and operate core runtime services for multi‑turn orchestration, tool calling, model routing, memory, streaming, and safety
• You’ll work across distributed systems, production observability, and ML infra integrations to deliver an experience that feels instant, accurate, and trustworthy — while optimizing cost and reliability
• Own impactful runtime problems end‑to‑end — from architecture and design to production launch and ongoing reliability
• Build and evolve core services for session lifecycle, streaming responses (e.g., gRPC/WebSockets), structured tool execution, memory/state, and policy/guardrails
• Design for performance, correctness, and cost: reduce p50/p95 latency, improve tail behavior, and optimize token/tool budgets
• Integrate with leading LLM providers (e.g., OpenAI, Anthropic, Google Gemini) and internal evaluation frameworks to improve quality and predictability
• Harden the platform with fault isolation, retries, timeouts, circuit‑breaking, backpressure, and graceful degradation
• Instrument deep observability (tracing, metrics, logs) and create playbooks/SLOs for high availability and on‑call excellence
• Collaborate closely with product, quality, and application teams to prioritize the most impactful roadmap investments