Tech Lead – Agentic AI Platform

About Multiscale AIMultiscale AI builds agentic AI systems for semiconductor manufacturing. Our platform combines domain retrieval, automated analytical workflows and agent orchestration to help fab engineers move faster on complex data problems. Offices in Atlanta, San Francisco and Hyderabad.The RoleSMPO AI is our agentic orchestration platform: a chat-driven system that plans and executes AI agent work in isolated compute environments and streams results back to the user in real time. As Tech Lead you will own the technical health of this platform end to end, act as Architecture Owner for the team, and be the single technical reporting point for our Hyderabad product engineers.This is a hybrid role: roughly two-thirds hands-on system design and code, one-third technical leadership.What You Will OwnPlatform architecture end to end: how a user request becomes an agent run, how execution is orchestrated, and how results stream back in real time.Reliability for long-running agents in disposable environments: recover cleanly from failures without losing or duplicating work; checkpointing and durable agent state.The lifecycle and security model of the sandboxed compute environments agents execute in.The security and audit model for agents: role-based access control, authentication passthrough and impersonation, immutable audit trails, identity and provisioning flows that customer security teams will review.LLM and tool gateways: MCP servers, tool registration and credential management, so every agent action goes through a controlled, logged interface.The retrieval layer: embeddings, vector stores and enterprise document access tuned for fab and materials-science data.Observability: tracing, monitoring and the evidence that agents behaved as intended.Evaluation as a release gate: no AI-driven feature ships without an eval suite that proves it against known cases.Support for network-isolated and air-gapped customer deployments.Performance and scale targets: concurrency, run duration, startup latency.Technical LeadershipArchitecture Owner: final reviewer on changes that affect platform architecture, including AI-authored ones.Lead the Hyderabad product team: single point of technical reporting, owning workload planning and balance in coordination with the US product lead.Drive spec-first development; set the quality bar for AI-driven features; mentor engineers on agentic system design and on using AI coding tools (Claude Code, Cursor) safely.Partner with DevOps on CI/CD: release-candidate promotion, ArgoCD GitOps deployment and production quality gates.Own hiring for the India product team: rubrics, technical screening and shortlists.What We Are Looking For — Required5+ years of software engineering, including 2+ years in a technical-leadership or architecture-ownership capacity.Strong backend engineering in Python; production services on databases and caching/queueing infrastructure (MongoDB and Redis a plus).Solid distributed-systems fundamentals: coordination, failure detection, idempotent operations.Experience building or operating systems that run workloads in containerised or ephemeral compute.Practical experience building LLM-powered or agentic systems: multi-step execution, streaming output, human-in-the-loop approvals (LangGraph or comparable).Hands-on experience with enterprise security patterns for AI systems: RBAC, authentication passthrough, audit logging; SOC 2 exposure a plus.Experience with the Model Context Protocol (MCP) or building tool-calling interfaces for agents.Experience defining and running eval suites for LLM-driven features and using them as release gates.Working knowledge of RAG pipelines end to end: embeddings, retrieval, evaluation, and where they fail.Experience with observability for distributed or agentic systems: tracing, monitoring, alerting.Experience leading a small engineering team day to day: planning, prioritisation, workload balance.Comfort with event-driven architectures; a track record of owning architecture decisions and reviewing others’ designs at a senior level.PreferredAI-assisted engineering workflows at team level (Claude Code, Cursor): project configuration, shared prompt libraries, spec-driven development.GitOps deployment (ArgoCD) and progressive delivery (canary / blue-green).Network-isolated or air-gapped deployment environments.Domain exposure to semiconductor manufacturing, industrial analytics or other data-heavy enterprise verticals.Why Join UsReal ownership from day one of a platform already in production for a Fortune-500 manufacturer.Small team, fast decisions, direct access to customers and their hardest problems.Competitive compensation.