Latest article
A deep agent anatomy
The four components behind deep agents — a detailed system prompt, a planning tool, sub-agents and a file system — and how they differ from shallow ReAct agents.
Ravi Shah
Executive Architect · IBM Consulting UK&I · London
I build software with AI agents and ship agentic capability into production for Financial Services — and write about what works: agentic patterns, context engineering, token budgets and security.
Writing
All articles →Token Budget Discipline: A Developer’s Playbook
A practical playbook for treating model quotas like engineering budgets: route requests, reserve premium capacity, and avoid cooldown stalls.
IBM Project Bob Shifts Security Left with AI-Powered Code Scans
How AI-powered Gitleaks scanning moves security from a pipeline gate into the developer's PR loop — shift-left security for the agentic era.
Patterns for Context Engineering in Agentic Applications
Architectural patterns for supplying and structuring context so production AI agents behave reliably at scale.
Open source
All projects →deep-agent-anatomy
Runnable companion to “A deep agent anatomy”: shallow vs. deep agents, the four components, and a from-scratch deep agent you can run with no API key.
gitleaks-repo-scan-agent
Shift-left security for the agentic era — an AI agent skill that runs Gitleaks in Cursor, Copilot CLI and CI, interprets findings and guides remediation before secrets reach production.
llm-token-budget
Utilities to keep LLM token spend under control. Companion code for “Token Budget Discipline”.
Speaking
Invite me to speak →- 2026IBM UK&I Chief Architects — Engineering culture
- 2024AWS European FSI Industry Analyst Summit — Financial Services