Product Manager Agentic AI

About the role

What you'll do

• Innefu Labs is looking for a Product Manager who has personally taken an agentic AI product from a blank whiteboard to a production system — not someone who has only added AI features to an existing roadmap. You will own the vision, architecture-level tradeoffs, and go-to-market for a multi-agent, LLM-driven product line, working shoulder-to-shoulder with engineering, applied AI, and design to ship autonomous systems that reason, plan, and act with minimal human handholding.
• This is a builder's role. You are expected to be technically fluent enough to sit inside architecture discussions on agent orchestration, retrieval pipelines, and evaluation frameworks — and translate that into a roadmap, prioritization model, and business outcome the rest of the company can rally behind.
• Location Delhi
• 5 Days Work from office
• What You Will Own
• Define and own the product vision, strategy, and roadmap for an agentic AI product built from zero — problem discovery, architecture direction, MVP scoping, and scale-up.
• Translate ambiguous, open-ended problems into a phased agentic system design: which tasks are agent-driven vs. deterministic, where human-in-the-loop checkpoints belong, and how autonomy expands release over release.
• Partner with engineering on core architecture decisions — agent orchestration and planning layers, tool-calling/function-calling design, memory and state management, RAG and retrieval pipelines, and multi-agent coordination patterns.
• Own the evaluation framework for agent quality — task success rate, hallucination/error rate, latency, cost-per-task, and safety guardrails — and use it to drive prioritization, not just measure it after the fact.
• Define prompt, tool, and knowledge-source specifications working directly with applied AI/ML engineers; review agent behavior transcripts and failure cases as part of the regular product cycle.
• Run structured discovery with enterprise customers and internal stakeholders to identify high-value workflows for autonomous or semi-autonomous automation.
• Own the build-vs-buy and framework decisions in collaboration with engineering (agent frameworks, vector databases, LLM providers, orchestration layers) and stay current on the fast-moving agentic AI ecosystem.

What you'll bring

• 8–12 years in product management, with demonstrable ownership of at least one AI/ML or agentic AI product taken from concept/zero to a live, adopted release — not just a feature add-on to an existing product.
• Working technical fluency in agentic AI system design: agent orchestration and planning (e.g., ReAct-style reasoning, task decomposition, multi-agent coordination), tool/function calling, and agent memory patterns.
• Solid working knowledge of the surrounding AI stack: LLM APIs and fine-tuning/prompting tradeoffs, RAG pipelines, vector databases, knowledge graphs, and evaluation/observability tooling for LLM-based systems.
• Familiarity with common agent frameworks and protocols (e.g., LangGraph, AutoGen, CrewAI-style orchestration, or equivalent in-house frameworks) and emerging standards such as MCP (Model Context Protocol) for tool/agent interoperability.
• Track record of writing clear PRDs/specs for ML-driven features, defining success metrics for probabilistic (not just deterministic) systems, and running structured experimentation.
• Comfort partnering closely with engineering and applied AI/ML teams on architecture-level tradeoffs — you don't need to write production code, but you need to hold your own in a systems design conversation.
• Strong stakeholder management and communication skills — able to translate deep technical tradeoffs into business language for leadership, sales, and customers, and vice versa.
• Prior experience in enterprise B2B SaaS, security/intelligence, fintech, or another data-sensitive domain is a plus, given Innefu Labs' customer base.
• Bachelor's degree in Computer Science, Engineering, or a related technical field; an MBA or equivalent is a plus but not a substitute for hands-on technical depth.

Nice to have

• Direct experience owning or contributing to a knowledge-graph or retrieval-heavy product (e.g., Neo4j, Qdrant/pgvector or similar vector stores).
• Exposure to agent safety/guardrail design — RBAC for agent actions, audit logging, human-in-the-loop escalation design.
• Past experience as an engineer, applied scientist, or technical founder before moving into product.