Siftstack Software Engineer, Backend (Agentic AI) at Siftstack develops backend systems for AI agents that analyze telemetry data for hardware engineers. The role involves designing scalable distributed systems, APIs, and evaluation tools.
Responsibilities
Hardware programs generate far more telemetry than any team can review by hand. Sift Agents is our answer: AI agents that work alongside hardware engineers, investigating anomalies, reviewing test and flight data, and answering in minutes questions that used to take a day of scripting.
This role comes with real product ownership and autonomy. You will talk directly to customers to understand how they review their machines, decide what to build, and own it from first MVP through every iteration after it ships. The surface area is wide: one day you might be designing the tool interface an agent uses to query billions of telemetry points, the next you might be writing the Kubernetes scaling logic that orchestrates agent workloads. We’re heavy users of frontier agentic products ourselves and have strong opinions about what makes them great. Sift Agents is where we put those opinions to work.
This is a product engineering role, not a research role. Most of the work is the hard part of making agents dependable: tool design, context management, code execution environments, evaluation, and observability.
Talk directly to customers and partner with product to turn real review workflows into agent capabilities: generating dashboards, writing analysis scripts, and surfacing insights buried in their telemetry
Design, ship, and operate agentic systems that reason over large-scale time-series data and hardware domain context
Build everything around the model that makes agents dependable: tool interfaces, sandboxed execution for agent-generated code, context and memory management, custom compaction algorithms, opinionated skills, and guardrails
Build the distributed systems that let long-running agent work stream in real time, survive disconnects, and resume across restarts
Design job-based execution systems that schedule, scale, and tear down agent workloads on Kubernetes, both in the cloud and on-prem
Develop and maintain Sift’s MCP server, the tool surface that lets both our agents and our customers’ AI tools query telemetry directly
Design and implement the APIs that power our agentic capabilities
Build evaluation suites that measure whether agents actually help engineers, and instrument quality, latency, cost, and failure modes in production
Integrate and assess frontier models across providers
Qualification
Experience building APIs (RESTShipped products to users at scaleShipped LLM-powered featuresBuilt agentic systemsDesigned tool ecosystems for agentsA personal ecosystem of AI dev tooling
Required
3+ years of professional software engineering experience
Experience building APIs (REST, gRPC)
Preferred
Shipped products to users at scale: large data volumes, significant active user counts, or deep technical complexity
Shipped LLM-powered features
Built agentic systems: multi-step tool use, planning loops, context management, and evals
Designed tool ecosystems for agents, including MCP
Worked with sandboxed or isolated execution of generated code
Operated services in production (Kubernetes, observability, incident response)
Worked with streaming or time-series data systems (Kafka, Flink, TimescaleDB)
A personal ecosystem of AI dev tooling: custom agents, skills, scripts, or workflows built to ship faster
