Principal AI engineer at Smartsheet in Bangalore (hybrid): owning the target-state architecture for the company's AI platform - agent harness, knowledge graphs and the shared skills library.

This Bangalore role was posted on 18 September 2026. This is a true principal-level architecture seat: you define the standards that bind data, agents and knowledge across Smartsheet's AI platform, architect the agent harness and sub-agent framework, and build the reference agent framework other teams fork. The listing asks for deep hands-on experience architecting production agentic systems on Databricks, AWS Bedrock/AgentCore and MCP.

About Smartsheet

Smartsheet is the Intelligent Work Management Platform: for over 20 years it has helped teams plan, track and automate work, and it is now uniting people, data and AI - including AI agents - in one platform. Its India Global Capability Center in Bangalore is described as leading global product innovation, not just supporting operations. Learn more on Smartsheet's website.

The role in plain terms

You own the AI platform's target-state architecture: agent orchestration, lifecycle and hand-off contracts; knowledge and context graph systems - ontology design, ingestion pipelines, retrieval and grounding; and a shared, versioned skills and tools library with clear interface contracts. The listing wants a builder's track record: rapid prototyping, shipped software and adoption-driven patterns rather than mandate.

What you will do

Per the employer's listing, you will:

Own the target-state architecture for Smartsheet's AI platform, defining the standards that bind Data, Agents, Knowledge Graph, Context Graph into one coherent system.
Set the architecture for the agent harness and sub-agent framework including orchestration, lifecycle, hand-off contracts, Evals (Offline and Online) and capability-tier classification used across all agent teams.
Design and own schema/ontology, ingestion and freshness pipelines, and the retrieval and grounding patterns that keep agent responses reliable.
Build and maintain a reference agent framework that teams fork or extend, so no team re-solves orchestration, tracing, or deployment from scratch.
Author and publish a shared skills and tools library with clear interface contracts and a versioning model any team can build on.

What the employer asks for

These requirements are quoted from the employer's listing, kept in their exact words so eligibility details are not distorted:

Deep, hands-on experience architecting production agentic systems on Databricks Platform, AWS Bedrock/AgentCore and MCPs
Experience in Agent harness/orchestration design, sub-agent boundaries, tracing and observability (e.g., MLflow), Evals, Governance, Responsible AI and cost/latency instrumentation at platform scale.
Strong background in Knowledge and Context graph based systems, including ontology design, ingestion pipelines, and retrieval/grounding evaluation.
A builder's track record: you prototype rapidly, ship working software, and let adoption — not mandate — prove which patterns are right.
Experience designing reusable, versioned software interfaces (skills, tools, libraries) that multiple teams adopt directly, with clear contracts and promotion criteria from prototype to production.
Familiarity with modern LLM/agent tooling and standards, including Databricks, Bedrock, Strands SDK, and MCP (Model Context Protocol).

Who this role suits

Staff-plus engineers who have actually shipped agentic systems in production - harness design, tracing, evals and grounding, not demos. The Databricks + Bedrock/AgentCore + MCP stack is specific; hands-on with at least part of it is needed. Not for ML generalists or engineers new to agent architectures.

Practical notes

This is a genuinely senior AI platform architecture seat. Expect deep architecture interviews: bring a production agentic system you designed and be ready to defend orchestration, eval and grounding choices. Hybrid in Bangalore per the listing.

How to apply

Apply directly on the employer's site: Principal AI Engineer at Smartsheet. We found this listing on the Greenhouse public job board (job-boards.greenhouse.io), where it was posted on 18 Sep 2026 and was still open when we checked on 31 Oct 2026. Before applying, re-read the employer's own description in full, since eligibility details, deadlines and process steps can change at any time and the employer's page is always the final word. Tailor your resume to the specific requirements above rather than sending a generic one: mirror the skills and tools the listing names where you genuinely have them. A short, specific note on why this team and this problem interest you goes further than a long generic cover letter. Never pay a fee to apply or interview; genuine employers, including this one, do not charge candidates.

Related openings on GetFirstJob

If this role is not the right fit, these current openings are in a similar area:

We write and maintain this page from the employer's public listing and link straight to their application page. We do not act as a recruiter, and applying is always free. If the link above stops working, the role has likely been filled.

This is our short summary of the employer's listing. The full description, eligibility and deadline are on the employer's page.