Efficient Prefetching Technique for Hadoop Distributed File System Federation
International Conference on Advanced Computing · IEEE
Dec 2013
The architecture call, the build nobody has scoped yet, the production problem three teams have already bounced off — I come in, figure it out, and own it end to end. Agentic architectures, AI-native runtimes, distributed systems, and ML infrastructure.
Modular AI-native runtime and company-brain system for enterprise business context. Pivoted from Malleable.
Graph-engineered agentic system researching product, churn, competitors, branding, and features — live in production, including for Tangible.
RL-based ML systems, petabyte-scale streaming data infrastructure, and backend services at scale.
Founded Malleable, an AI-native runtime that enables coding agents to make targeted, context-aware edits instead of regenerating entire applications — reducing disruption while supporting continuously evolving software. Validated the approach through pilots with ~500 users across two university campuses.
Reimagined enterprise software as an AI operating layer, founding Operable — a runtime that gives AI agents governed business context spanning customers, contracts, invoices, and relationships instead of isolated APIs. Currently piloting with enterprise customers.
Delivered a production agentic research platform for an AI company by combining knowledge graphs, persistent memory, planning, and execution into a reliable multi-agent system that continuously researches products, competitors, and market trends.
Built reinforcement learning systems at Google to estimate online-to-offline store visits by training and serving models over petabyte-scale datasets combining location history with advertising events.
Designed distributed streaming infrastructure processing petabyte-scale data, and built self-healing observability systems capable of monitoring and automatically updating production Dataflow pipelines.
Launched Google Cloud Dataflow Flex Templates, driving 10× adoption within three months while simplifying deployment for enterprise customers.
Led products from zero to one — owning customer discovery, product strategy, architecture, engineering execution, and enterprise pilots while iterating through multiple pivots based on market feedback.
I get back within 24 hours.
Happy to jump on a call — I'm on Pacific time. WhatsApp or a text works just as well.
I take on a small number of independent engagements alongside Operable — agentic systems, AI-native runtimes, and ML & data infrastructure. Architecture reviews, zero-to-one builds, or getting an AI system that demos well actually running in production. Tell me what you're building.