Soham Shah
Designing the interface between human intent and machine autonomy.
Building agent systems that compress context, coordinate specialists, and sustain coherence across long-horizon tasks.
Active / SF Bay / Agent Architecture
Research
Self-Evolving Skills in Multi-Agent Systems
A survey of the Feb–April 2026 explosion: how skills became the npm of agentic AI — versioned, composable, marketplace-distributed, and autonomously evolving.
researchagent-skillsmulti-agent
Research
Context Compaction for Long-Running Agents
Agent sessions degrade as context grows -- not at the limit, but continuously. A research-grounded design for keeping context lean and high-signal, synthesizing the latest work from 40+ papers.
researchcontext-engineeringagent-systems
Architecture
Specialized Subagent Architecture
From generic delegation to context-aware multi-agent coordination -- how to decompose, delegate, persist, and train specialized agents.
architecturemulti-agentsystems-design