AI Development Engineer
Cywarden · AI Agentic Cyber Platform · Engineering · 12-week contract to hire
Location: Bay Area (Dublin, CA) · work from office
About Cywarden Cywarden is a cybersecurity and technology firm that advises, implements and runs 24x7 security and IT operations for enterprises, with delivery centers in Pune and Mohali. We partner with Google Cloud, Microsoft Sentinel, Wazuh and SolarWinds. We are building the Cywarden AI Agentic Cyber Platform. Enterprises deploy it inside their own Azure or Google Cloud tenant, and its AI agents do real security work: triaging alerts, managing vulnerabilities, running pentests, reviewing firewalls and pushing approved fixes. Every agent is backed by Cywarden's security engineers and 24x7 operations.
The role:
You will build the agent runtime that powers every module of the platform. Our rule is simple: code does the math, the AI explains. Deterministic engines find, score and correlate; the language model reasons about intent, writes the explanation and drafts the fix. You make that split reliable, safe and affordable at enterprise scale. What you will build The agent runtime: planning, tool calling, memory and state for agents such as Firewall Review, Alert Navigator and the AI pentest agent. MCP servers and tools that let agents read from Sentinel, Google SecOps, Tenable, Service Now, GitHub and cloud APIs, read-only by default. Guardrails: approval gates before any action, scoped credentials, prompt-injection defenses, and refusal of out-of-scope steps. Evaluation suites that score each agent's verdicts, explanations and drafted fixes against expert-labelled cases, and run on every release. Model routing across Gemini on Vertex AI and Azure OpenAI, with cost, latency and quality budgets per task. The Agent Studio back end, so security teams can build agents in plain language with the platform checking the logic. Observability: traces of every agent decision, so a client's auditor can see what the agent saw and why it acted. Tools you will use Python, TypeScript, Vertex AI (Gemini), Azure OpenAI, MCP, LangGraph or similar, Postgres, Pub/Sub and Service Bus, Open Telemetry, GitHub Actions. What you bring 5+ years of software engineering, including 2+ years shipping LLM features to production. Strong Python or TypeScript; clean APIs, tests and code reviews. Hands-on tool calling and agent orchestration, and a clear view of when not to use an LLM.Experience building or consuming MCP servers, or similar tool protocols. A habit of measuring: evaluation sets, regression tests and A/B comparisons for prompts and models. Working knowledge of Vertex AI or Azure OpenAI, including quotas, private endpoints and data residency. Clear written English; you will explain agent behavior to clients and auditors.
Nice to have:
Security background: SOC, detection engineering, vulnerability management or penetration testing.
Retrieval and embeddings over security data; fine-tuning or distillation for cost.
Experience with Claude, Gemini and GPT model families side by side.
Contributions to open-source agent or MCP projects.
How to apply:
Send your resume and a short note on an agent or LLM system you shipped, what went wrong, and how you measured it. Cywarden is an equal opportunity employer. We welcome applicants of every background.