Associate Principal Software Engineer, AI Onboarding in San Francisco, California at Saviynt

Part of the team to build the Onboarding and AI Platform. The team builds an AI-based platform and onboarding experience for customers.

• You will build the Pilot Rail (the human-in-the-loop approval interface for AI-generated plans), execution wave UIs (step-by-step progress tracking for multi-day agent workflows), and the policy gap reporting surface.
• Also will be responsible for connector health APIs, IaC template generation endpoints, plan envelope management, and the approval chain service.
• Across various products, you will contribute to the shared platform layer: the signed context packet format (the structured memory object passed between the human UI and the agent), the plan state machine (the lifecycle model for AI-generated plans from draft through approval to execution), the Pilot agent panel (the contextual guidance component embedded in every wizard and workflow), and the audit log service.
• Work across multiple agent platforms like AWS Bedrock, Google AgentSpace, Salesforce AgentForce, building foundational solutions, using cloud, SAAS and AI design patterns and technologies.
• Use AI and Agents to secure AI, using CUA agents, various LLM’s, agentic frameworks like ADK, Langchain, among others.
• Design and develop secure, scalable, multi-tenant software solutions that run seamlessly across major cloud platforms like AWS, Azure, and GCP.
• You will design and build product features end-to-end, from the React component through the API route to the database schema.
• You will write clear, testable code and participate in code review with the expectation that your feedback improves the codebase, not just the PR.
• You will work directly with product managers and designers to refine requirements before writing a line of code, and you will flag ambiguities early rather than building on an assumption.
• You will be responsible for the quality of your own work. That means writing unit and integration tests, monitoring the services you deploy, and fixing the bugs you introduce. It also means reading error logs, understanding the failure modes of the systems you build, and designing for graceful degradation — all of which are particularly important in an agentic context where the AI layer may be unavailable or return unexpected results.
• You will contribute to the team's engineering standards: API design conventions, component patterns, state management approaches, and the shared vocabulary used across the codebase. You are not expected to define these standards alone, but you are expected to engage with them seriously.
• At 3–5 years, we expect you to own a feature end-to-end with light guidance. You ask good questions before starting, surface blockers early, and deliver working software that meets the acceptance criteria. You write code that your teammates can read without a walkthrough. You fix your own bugs. You participate in design discussions and share your opinions, but you update them as you learn new information.
• We do not expect you to architect the system, lead the team, or have opinions on every technology choice. We do expect you to be curious, to read the codebase before asking questions it can answer, and to take pride in the quality of what you ship.

Required:

• You have 1-2 years of Staff or Associate Principal level professional software engineering experience, with at least one year building production web applications. You are comfortable working across the full stack and have a clear preference for writing code that other engineers can read and maintain.
• Working with LLM APIs (OpenAI, Anthropic, Amazon Bedrock) or agent frameworks (LangChain, LangGraph, or similar).
• You have experience building REST APIs in Python or a comparable language. You understand HTTP semantics, JWT authentication, and the basics of relational database design.
• You have written SQL queries, understand indexing at a conceptual level, and know when to reach for an ORM versus raw SQL.
• You have worked with Git in a team environment — branching, pull requests, code review — and you understand why these practices exist.
• You have deployed code to a cloud environment (AWS, Azure, or GCP) at least once and have a basic understanding of what happens between `git push` and a user seeing your change.

Nice to have

• On the frontend, you have solid working knowledge of React and TypeScript. You understand the component lifecycle, state management patterns, and the differences among local state, context, and server state.
• Understanding of the human-in-the-loop pattern in agentic systems — specifically, how to design approval gates that are substantive rather than ceremonial.
• You have built forms, tables, and multi-step workflows. You are familiar with REST API integration from the browser and know how to handle loading states, error states, and stale data gracefully.
• Experience with WebSocket or Server-Sent Events for real-time UI updates. Familiarity with identity and access management concepts: SSO, SAML, SCIM, RBAC, JIT access, privileged access management.
• Experience with FastAPI, SQLAlchemy, or PostgreSQL. Exposure to IaC tooling (Terraform, AWS CDK, ARM templates) from a developer perspective — you do not need to be an infrastructure engineer, but understanding what these templates do is useful context for the PAM work.