Jawad Amin
Customer Engineer at Google | AI Infrastructure
Welcome to my personal site for AI hardware benchmarking, TPU recipes, and engineering notes.
Blogs & Notes
- [July 16, 2026] — Cutting Coding Agent Costs by 15X: Serving Qwen 3.6 27B on Cloud TPUs
Notes, setup recipes, and commentary on serving Qwen 3.6 27B on Cloud TPUs and integrating with an agentic harness.
Public Repositories
- qwen3.6-27B-tpu-pi-agent
Benchmarking infrastructure and setup recipes for serving Qwen 3.6 27B on Cloud TPUs. - tpu-recipes
Collection of Google Cloud TPU optimization recipes and benchmarks.
View all public Google repositories on GitHub ↗
Past Work (AI Accelerators & Agentic Systems)
- WorkbenchIQ
Enterprise solution accelerator providing a modern workbench for underwriters and claims processors, combining AI Content Understanding and AI Foundry capabilities. - agentic-diligence-ibanking
AI-first agentic M&A due diligence platform powered by the Copilot SDK. - agentic-rcsa
Intelligent Risk and Control Self Assessment Accelerator with multi-agent architecture leveraging the OpenAI Agents SDK. - sage-retirement-planning
Personalized retirement scenario planning, financial projections, and AI-powered recommendations. - mike-ai-agent
Speech-to-Action AI agent implementation. - credit-card-agentic
AI-powered application system built with modern agent frameworks for autonomous multi-agent approvals. - document_indexer_gpt
PDF data extraction application using OpenAI vision models.