Senior Software Engineer, Backend (Agentic AI)

Sift Stack, Inc.•San Francisco, CA

14h•$180,000 - $230,000•Hybrid

About The Position

Sift is hiring a Senior Software Engineer, Backend (Agentic AI) to build Sift Agents, AI agents that help hardware engineering teams investigate anomalies, review test and flight data, and analyze high-frequency telemetry. This is a product engineering role focused on making agents dependable, rather than a research role. The position has product ownership and autonomy, including working directly with customers to understand their workflows, deciding what to build, and owning features from MVP through subsequent iterations. The role spans agent tools and interfaces, context and memory management, sandboxed code execution, evaluations and observability, distributed systems, Kubernetes-based workload orchestration, and APIs. The engineer will also develop Sift’s MCP server and assess and integrate frontier models across providers. Sift’s relevant technologies include Go, gRPC, PostgreSQL, Rust, Flink, Kubernetes, AWS, Grafana, Prometheus, and others. Sift is headquartered in Marina Del Rey, California. The team collaborates in person on Mondays and Thursdays and meets for a full week every two months. Relocation to Los Angeles or working from the San Francisco office may be possible for the right candidate. The stated salary range is $180,000–$230,000 per year, plus equity and benefits. Applicants must be a U.S. citizen, lawful permanent resident, or protected individual such as an asylee or refugee, in compliance with ITAR/EAR regulations.

Requirements

At least 8 years of professional software engineering experience.
Interest in owning a product area, including talking to customers, deciding what to build, and shipping it.
Experience building APIs such as REST or gRPC and complex backend services using Go, Python, Rust, or similar technologies.
Working knowledge of distributed systems fundamentals.
Curiosity about new AI products, with experience trying new agents, models, and features and opinions about what makes them effective.
Must be a U.S. citizen, lawful permanent resident, or protected individual such as an asylee or refugee, in compliance with ITAR/EAR regulations.

Nice To Haves

Shipped products to users at scale, involving large data volumes, significant active user counts, or deep technical complexity.
Shipped LLM-powered features.
Built agentic systems involving multi-step tool use, planning loops, context management, and evaluations.
Designed tool ecosystems for agents, including MCP.
Worked with sandboxed or isolated execution of generated code.
Operated production services, including Kubernetes, observability, or incident response.
Worked with streaming or time-series data systems such as Kafka, Flink, or TimescaleDB.
Built a personal ecosystem of AI development tools, such as custom agents, skills, scripts, or workflows.
Background in time-series data, scientific computing, or hardware test and telemetry.
Built internal agentic tooling that accelerates an engineering organization.

Responsibilities

Work directly with customers and product to turn real review workflows into agent capabilities, such as generating dashboards, writing analysis scripts, and surfacing insights in telemetry.
Design, ship, and operate agentic systems that reason over large-scale time-series data and hardware-domain context.
Build systems that make agents dependable, including tool interfaces, sandboxed execution for agent-generated code, context and memory management, custom compaction algorithms, skills, and guardrails.
Build distributed systems that stream long-running agent work in real time, survive disconnects, and resume across restarts.
Design job-based execution systems to schedule, scale, and tear down agent workloads on Kubernetes in cloud and on-premises environments.
Develop and maintain Sift’s MCP server so agents and customers’ AI tools can query telemetry directly.
Design and implement APIs for agentic capabilities.
Build evaluation suites and instrument agent quality, latency, cost, and production failure modes.
Integrate and assess frontier models across providers.