Senior AI Engineer
Zafran · Tel Aviv · onsite
Zafran is building the next generation of vulnerability and risk management. We bring together complex security data and turn it into clear investigation, prioritization, and action.
We're creating a new AI team to build the shared infrastructure, tools, and standards that enable teams across R&D to develop reliable AI-powered capabilities. The team will also own and deliver core AI experiences used across Zafran's platform and organization.
About Zafran:
Our Mission: To stop the exploitation of vulnerabilities, everywhere.
What makes us different: Zafran de-risks 90% of critical vulnerabilities overnight across your hybrid environment, leveraging Agentic Capabilities and your existing security tools to rapidly mitigate and remediate the 10% most likely to be exploited.
Who’s behind us: Zafran is backed by Menlo Ventures, Sequoia Capital, Cyberstarts, and a deep belief that cybersecurity should move as fast as attackers do. We’re one of the fastest-growing companies in the industry, scaling to meet demand from the world’s most advanced, security-obsessed organizations.
We’re serious about our mission- so expect work that matters, teammates who challenge and inspire you, and plenty of fun along the way!
We're looking for a Senior AI Engineer to be one of the team's first engineers and to own Zafran's AI infrastructure end to end — the shared platform every R&D team uses to build, test, ship, and monitor AI features. That platform powers our in-product AI assistant, our background agents, and every AI capability still to come. You'll set the standard for how AI runs in production at Zafran: when a team ships an AI feature, it should be reliable, secure, observable, and cost-efficient by default — because your platform made it that way.
What you will do:
- Own the AI platform. Design, build, and run the shared agent runtime: model access, tool use, memory and sessions, streaming, and human approval for sensitive actions.
- Production-grade. Scaling, fallback between model providers, rate limits, timeouts, and safe rollouts on Kubernetes.
- Own evaluation. Build the evaluation framework every team uses to measure quality, correctness, latency, and cost — and catch regressions in CI before customers do.
- Own observability and cost. Tracing for every agent run and tool call, token spend per tenant and per feature, and alerts when quality or cost drifts.
- Own AI security. Strict tenant isolation, defenses against prompt injection, and guarded data access across a multi-tenant security platform.
- Build the tool layer. MCP tools and data connectors that let agents query our security data safely and reliably.
- Enable R&D. Shared libraries, templates, and docs so other teams can build, test, deploy, and monitor AI features independently.
- Stay hands-on with architecture, implementation, and production issues, and help raise the AI engineering bar across R&D.
Requirements:
- 5+ years of backend or infrastructure engineering, with strong system design skills.
- Proven experience building and operating production-grade AI / LLM infrastructure — real users, real traffic, real incidents. Not prototypes.
- Deep hands-on experience with agentic systems: tool use, structured outputs, prompt and context management, retrieval, and long-running or streaming flows.
- Experience designing and running LLM evaluations at scale.
- Strong Python skills.
- Production experience on AWS and Kubernetes, including observability (tracing, metrics, logs).
- Ownership mindset — you treat the platform as a product and other engineers as your customers.
Experience with the following is a plus:
- Experience with Amazon Bedrock, Strands Agents, LiteLLM, MCP, or OpenTelemetry (our current stack).
- Experience building internal developer infrastructure used by multiple engineering teams.
- Experience with model evaluation, safety controls, governance, or LLM security (guardrails, prompt injection).
- Experience operating multi-tenant SaaS systems.
- Background in cybersecurity or security-data products.
- Familiarity with Go, ElasticSearch, ClickHouse, or Postgres.
Why this role matters:
AI is a shared capability across Zafran's products and engineering organization. As the owner of the AI infrastructure, you'll build the foundation that lets every R&D team ship AI features faster and more safely — and shape Zafran's AI architecture from the ground up.