Why Kubiya AI?

Kubiya is building the agentic operating system for engineering, cloud and DevOps teams.

Our platform stitches together open-source LLMs, a Mixture-of-Experts runtime, infrastructure-native tooling (Kubernetes, Terraform, OTEL, JetStream) and a policy-first security layer to turn a single prompt into a fully-auditable, production-ready workflow. Fortune-100 design partners already collapse 300-step runbooks into “spin up an environment” conversations — no glue code, no pager fatigue.

⸻

The Opportunity

As AI Team Lead in the CTO’s office, you will be the technical owner of everything model-powered inside Kubiya:

• Architecture – Design the end-to-end pipeline that ingests org context, routes to the right expert model, executes code in sandboxed containers, and feeds rich telemetry back into our continuous-learning loop.

• Model Strategy – Decide when we fine-tune open-source Llama-3 vs. hot-swap to Bedrock or Vertex; benchmark MoE routers for latency and cost; champion vLLM/Triton for GPU efficiency.

• MLOps at Scale – Own versioning, lineage, policy gating and roll-back of models and in-line tools. Ship deterministic, reproducible releases that DevSecOps trusts.

• Tooling & Integrations – Work with backend and platform leads to expose new model endpoints through our Model Context Protocol (MCP) so agents can compose actions across GitHub, Jira, Terraform, Prometheus and more — without one-off plugins.

• Thought Leadership – Partner with the CTO on the technical roadmap, publish internal RFCs, mentor engineers and evangelize best practices across the company and open-source community.

⸻

What You’ll Do

• Craft cloud-native, micro-service architectures for training, fine-tuning and real-time inference (AWS/GCP/Azure, Kubernetes, JetStream).

• Define SLOs for p95 agent latency, model success rate, and telemetry coverage; instrument with OTEL, Prometheus and custom reward models.

• Drive our continuous-learning loop: reward modelling, ContextGraph enrichment, auto-tuning MoE routers.

• Embed least-privilege IAM and OPA/ABAC policy checks into every stage of the model lifecycle.

• Collaborate with product managers to translate customer pain into roadmap items and with design partners to validate solutions in production.

• Mentor a cross-functional squad of backend engineers, ML engineers and data scientists.

⸻

What You Bring

• 5+ years in software engineering, with 3+ years architecting large-scale backend systems (Python, Go, Java or similar).

• 4+ years designing, deploying and monitoring AI/ML systems in production.

• Deep expertise in at least one of: large-language-model serving, MoE routing, RLHF, vector search, streaming inference.

• Hands-on fluency with Kubernetes, Docker, CI/CD, IaC (Terraform/Helm) and distributed data technologies (Kafka, Spark, Arrow).

• Proven MLOps track record (MLflow, Kubeflow, SageMaker, or similar) and a security-first mindset.

• Ability to turn ambiguous business goals into a crisp, scalable architecture — and to communicate that vision to both executives and engineers.

⸻

Nice-to-Haves

• PhD or publications in ML/NLP/Systems.

• Contributions to open-source LLM or MLOps projects.

• Experience pushing real-time inference to the edge or FPGA/ASIC accelerators.

• Prior leadership of cross-functional AI/ML teams in a fast-growing startup environment.

⸻

The Way We Work

We value clarity, ownership, and velocity. You’ll have direct access to the CTO, autonomy to choose the right tech, and a front-row seat as we redefine how enterprises move “from prompt to production.” If building the Kubernetes of AI-driven operations excites you, let’s talk.