This title did not exist two years ago, because the way of working it describes did not exist. An agentic software architect builds agents using agents. Not one assistant autocompleting code — fleets of them: agents that write the code, agents that review it, agents that test it, agents that watch it run in production, all orchestrated by an architect who thinks less like a bricklayer and more like a conductor. Your output is not measured in lines you typed. It is measured in the systems your agents shipped — and in how many agents those systems can build in turn.

The craft goes all the way down the stack. Agentic systems: runtimes where dozens of agents plan, delegate, verify, and recover without a human in every loop. Agentic solutions: customer-facing deployments where document pipelines, decision workflows, and domain agents compose into something an enterprise trusts with real operations. Agentic search: retrieval that reasons — agents that decide what to look for, interrogate what they find, and know when the answer is not good enough yet. At Botminds all of this is not a lab experiment; it is the product, and this role owns its architecture.

What you’ll do

You will design and evolve the multi-agent architecture of the Botminds platform — how agents are defined, composed, orchestrated, evaluated, and trusted — and you will practice what you architect: running fleets of coding agents to build the platform itself, turning what works into the patterns, harnesses, and guardrails the whole team adopts. You set the standards for agent-built software: how agent output is verified, how multi-agent workflows recover from failure, how agentic search grounds every answer, and where the human stays in the loop because judgment demands it. You prototype at the frontier — new orchestration patterns, agent-to-agent protocols, self-improving pipelines — and you carry the best of it into production, where lending and credit operations depend on it being right.

What you bring

8+ years building production software, with deep recent hands-on work in LLM-based systems — orchestration, RAG, multi-agent frameworks, or serious agentic coding practice.
A demonstrated agent-native way of working: you already build with agents daily and think in delegation, verification, and feedback loops rather than keystrokes.
Architectural depth to make agentic systems safe for enterprises: evaluation, observability, failure recovery, and governance are design inputs for you, not afterthoughts.
The conviction to define a discipline — writing, teaching, and standard-setting are part of the job, because this role has no textbook yet.

Why this role

Because the leverage is unlike anything in software today. An architect who commands agents that build agents does not add to a team’s output — they multiply it, and the multiplier compounds. You will define what this title means while shipping the systems that prove it, on a platform where agentic AI already runs real institutions. In five years every serious company will have this role. Here, you get to invent it. Full time, Engineering department, based in Chennai.