“The journey that never ends”
AI systems, built to stay current
I design agentic systems that retrieve from private knowledge, refresh as the corpus moves, and stay operable in production. Notes from that work, and the path that led here.
Watch it search — depth-first, then breadth-first
Hyderabad, India
AI Architect @ Space · Enterprise AI at Scale
AI Architect at Space, designing enterprise AI platforms that turn messy organisational knowledge into reliable agentic systems.
Current role
Space
Jul 2026 — Current
An agent that can plan, call tools, and loop is easy to demo. An agent you can operate is a different artefact. Operable means: when it fails, you know which tool, which state, which prompt version, and whether a human should have been in…
The demo is easy: chunk a folder, embed it, ask a question, get a cited answer. The production problem is different. Enterprise knowledge is not a snapshot. Contracts get revised, runbooks drift, tickets close, wikis fork. If retrieval is…
Focus
Problems I am actively exploring — where models, agents, and graphs meet production constraints.
Task-specialised models that stay cheap, private, and deployable on real infrastructure — distillation, routing, and evaluation.
Agents that improve from feedback loops: tool use, memory, critique, and guarded online adaptation without drifting off-policy.
Knowledge graphs, hybrid retrieval, and graph-native planning for multi-hop reasoning over enterprise corpora.
Multi-agent retrieval stacks with continuous refresh, reranking, and faithfulness checks under terabyte-scale corpora.