About

Jigar Joshi — Agentic AI Architect & Founder

Agentic AI consultant, trainer & architect - helping teams ship production-grade systems.

Jigar trains development teams and consults with IT companies to build production-grade agentic AI systems. Every engagement is customised to the client's tech stack and produces working deliverables. He runs this through Wan Buffer, his agentic AI consulting and training firm (jigarjoshi.in is his personal site) — so when you read “we”, that is the Wan Buffer team of 20 people plus 150 agents in production.

Jigar Joshi, agentic AI consultant and founder of WanBuffer
15+
years in software, ERP & AI
8+
IT services teams trained
150+
agents in production
40%
faster delivery after training
Speaking & featured
Roles & leadership

Companies, ventures & how Jigar works with the industry

  • CEO & Founder of Wan Buffer Services
  • Agentic AI Architecture & Consultant
  • Agentic AI Trainer and Builder
  • Founder of ScanDoc Solutions
  • CEO & Founder of House of Aakrutii
Career timeline

15+ years, from IT services to agentic systems.

2007–2010
MCA, Computer Engineering
University of Mysore
2010–2018
Software & ERP / Odoo engineering
Roles across India and Kuwait
2018
Founded Wan Buffer Services
Ahmedabad, Gujarat
Today
CEO & Founder · Agentic AI architect, consultant & trainer
Wan Buffer Services — 20 people + 150 agents in production
The Work

Training teams. Building systems. Teaching builders.

Jigar runs WanBuffer, an Agentic AI consulting and training firm. He delivers corporate training to development teams, tech leads, and IT company leaders.

His training is known for being customised to the client's tech stack, using real client code, and producing working deliverables in every session. He has trained teams at 200OK Solutions, Bytes Technolab, and Metizsoft.

Founder, WanBuffer
Corporate Training for IT Companies
Agentic AI Implementation Consulting
Daily Content on AgenticAi Farming
Areas of Expertise
Agentic AI ArchitectureClaude API & Claude CodeModel Context Protocol (MCP)Multi-Agent SystemsRAG & CAGCursor IDEPrompt EngineeringProduction DeploymentTool Design PatternsLLM Observability
The Three Principles
Simplicity first
Every concept must be explainable without jargon. If it requires jargon to explain, the explanation is not ready.
Authenticity over polish
Write and teach from what has been built and observed. Specific details beat broad statements every time.
Build in every session
80% hands-on. 20% concept. Never reverse this. Every session ends with working code.
New on the blog: agent eval sandbox escape checklist
Anthropic's July 30 cyber-eval disclosure is a harness lesson. Prove egress, strip publish credentials, and match prompts to the real network before you raise autonomy.
Read the sandbox checklist
New on the blog: MCP 2026-07-28 day-one field notes
GA is live. Watch Session-Id leftovers, handle ownership under round-robin, MRTR clients, and cacheScope. Boring traffic is the win condition.
Read the day-one notes
New on the blog: Claude Code graph engineering, 14 steps
Linear agents are degenerate graphs. Dynamic workflows move orchestration into code: contracts, fan-out, verifiers, model tiers, and self-routing.
Read the roadmap
New on the blog: MCP portability when enterprise protocols compete
Mid-July reporting put five enterprise vendors behind a shared agent protocol framed against MCP. Finish the July 28 migration anyway, and keep tool contracts above the wire.
Read the playbook

Want to work with Jigar?

Training, consulting, or a discovery call to figure out what your team needs.

Get in touch