SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Ditto is a peer-to-peer sync platform that enables developers to build resilient, real-time applications regardless of network conditions. The company has raised $145M+ and is trusted by organizations like Chick-fil-A, Delta Airlines, and the U.S. military across aviation, retail, defense, and hospitality sectors.
As an AI Engineer, you will build Symphony, Ditto's agent execution platform that automates software engineering workflows. Rather than handcrafted changes, you'll create a system where AI agents handle development tasks while engineers oversee the process. This role bridges AI engineering and developer tooling—you'll own the complete stack from agent runtime to evaluation infrastructure.
Key responsibilities include: designing and shipping the agent execution runtime (topology, handoffs, deterministic gates, message bus, session lifecycle, recovery); building the typed tool surface that agents use to interact with systems safely (sandboxing, process supervision, MCP servers); establishing measurement and evaluation frameworks (evals, benchmarks, metrics, cost attribution); managing the model substrate as a swappable input (capability classes, routing, inference infrastructure); and partnering with Ditto's engineering teams to establish agentic development patterns.
You need 5+ years of professional software engineering with meaningful production experience building and operating LLM-backed systems (not prototypes). Required skills include hands-on depth with agentic systems (tool calling, multi-agent orchestration, prompt engineering), strong developer-tooling instincts from building CLIs, build systems, or CI/CD platforms, rigorous evaluation methodology, strong Python (async, typed interfaces, Pydantic), and systems fundamentals (processes, sandboxing, filesystems). You should have a track record of ownership, technical direction-setting, and working effectively in distributed teams.
Nice-to-haves include MCP experience, sandboxing/host isolation expertise, release engineering background, observability and cost accounting at scale (OpenTelemetry), inference infrastructure experience (vLLM, model routing), fine-tuning experience with open-weight models, and familiarity with Rust, Swift, or Kotlin.
