Google Developer Expert in AI/ML, Independent AI Consultant (Ex-Founder of Kanishka IT, 2010–2025), AI educator, and startup mentor. I help engineering teams build production-grade Agentic AI systems, optimize GenAI API costs, and execute AI roadmaps.
Most teams route every query to their most expensive model — GPT-4o, Gemini 2.5 Pro, Claude Opus — even for tasks a smaller model handles just as well. The result: 60–80% of your AI spend is waste.
Sending a simple classification task to a $15/million-token model when a $0.10 model does it equally well
No routing logic — every user query hits your most expensive endpoint regardless of complexity
Paying for cloud GPU compute when Google TPUs could run the same fine-tuning at 60–70% lower cost
Route low-complexity queries to Gemma / smaller models. Reserve expensive models only for tasks that genuinely need them. Typical result: 50–70% API cost reduction.
I review your current model choices, prompt design, and infra setup. You get a written report with exactly where your money is going and how to cut it — without losing quality.
Many use cases don't need a frontier model. I help you fine-tune and deploy open-weight models on Vertex AI or on-device — eliminating per-token API costs entirely for eligible workloads.
Semantic caching, prompt compression, and batching strategies that reduce redundant API calls — often cutting costs by an additional 20–30% with zero change to your product.
Geeta Kakrani · GDE AI/ML · AI-TPU Specialist · Bengaluru
Hands-on live workshops, developer bootcamps, and 1-on-1 mentorship sessions led by Geeta Kakrani to take you from API consumer to core systems engineer.
Learn how to design stateful multi-agent systems using LangGraph, CrewAI, and Google Antigravity. Understand routing, semantic caching, and memory systems.
Deep dive into Retrieval-Augmented Generation architectures. Learn document chunking, hybrid search, rerankers, and evaluations using RAGAS.
Custom 1-on-1 and team guidance for AI implementation, Vertex AI setup, GKE cost optimization, and open-source model deployments.
Verified public speaker and technical educator milestones across the global developer ecosystem.
Step-by-step guides, live coding, and open-source platform demonstrations.
Learn how to orchestrate multiple cooperative AI agents using Google's Agent Development Kit (ADK) in Hindi.
Learn how to run open-weights Gemma models locally with Ollama and orchestrate agents with the Google Agent Development Kit.
A beginner-friendly step-by-step tutorial on designing, writing, and executing your first autonomous AI Agent with Google ADK.
Join a dedicated network of developers, engineers, and machine learning practitioners. Discuss research, collaborate on open-source code, share ideas, and get architecture feedback.
Join WhatsApp Group