---
title: "Keshav Agarwal (@keshavagarwal) | Hashnode"
description: "Technical Founder & AI Product Leader | Real AI, real numbers, no leaderboard hype. Read the latest articles by Keshav Agarwal on Hashnode."
url: "https://hashnode.com/@keshavagarwal"
---

@keshavagarwal

# Keshav Agarwal

@keshavagarwal · Delhi, India · Joined December 2022

Technical Founder & AI Product Leader | Real AI, real numbers, no leaderboard hype

### [MCP, Tool Calling, and Intent-Based AI: What's Actually Different (and When It Matters)](https://beyondthebenchmark.hashnode.dev/mcp-tool-calling-and-intent-based-ai?utm_source=hashnode&utm_medium=feed)

No-MCP, tool calling, intent-based AI, and real MCP compared — what each actually means, and a clear framework for when to use which.

6 min read · [beyondthebenchmark.hashnode.dev](https://beyondthebenchmark.hashnode.dev/mcp-tool-calling-and-intent-based-ai?utm_source=hashnode&utm_medium=feed)

[Jun 28](https://hashnode.com/posts/mcp-tool-calling-and-intent-based-ai/6a40c794b2ea19090dda6c06)

### [Why Agentic AI Needs an Evaluation Stack (and Most AI Products Don't Have One)](https://beyondthebenchmark.hashnode.dev/why-agentic-ai-needs-an-evaluation-stack-and-most-ai-products-don-t-have-one?utm_source=hashnode&utm_medium=feed)

Why agentic AI breaks the eyeball test, and the eval stack (RAGAS, Promptfoo, Langfuse) that catches what a quick skim misses.

7 min read · [beyondthebenchmark.hashnode.dev](https://beyondthebenchmark.hashnode.dev/why-agentic-ai-needs-an-evaluation-stack-and-most-ai-products-don-t-have-one?utm_source=hashnode&utm_medium=feed)

[Jun 24](https://hashnode.com/posts/why-agentic-ai-needs-an-evaluation-stack-and-most-ai-products-don-t-have-one/6a3b70ff2eb16187e282337a)

### [How I Actually Choose an AI Model for Production (Not From a Leaderboard)](https://beyondthebenchmark.hashnode.dev/how-i-actually-choose-an-ai-model-for-production-not-from-a-leaderboard?utm_source=hashnode&utm_medium=feed)

Why I treat AI model selection as a product decision, not a leaderboard chase, and what shipped it to production.

5 min read · [beyondthebenchmark.hashnode.dev](https://beyondthebenchmark.hashnode.dev/how-i-actually-choose-an-ai-model-for-production-not-from-a-leaderboard?utm_source=hashnode&utm_medium=feed)

[Jun 19](https://hashnode.com/posts/how-i-actually-choose-an-ai-model-for-production-not-from-a-leaderboard/6a3556882d502689bdc75300)

### [🔥 Apache Kafka with Zookeeper in Django – A Complete Guide](https://beyondthebenchmark.hashnode.dev/apache-kafka-with-zookeeper-in-django-a-complete-guide?utm_source=hashnode&utm_medium=feed)

Apache Kafka with Zookeeper in Django – A Complete Guide

4 min read · [beyondthebenchmark.hashnode.dev](https://beyondthebenchmark.hashnode.dev/apache-kafka-with-zookeeper-in-django-a-complete-guide?utm_source=hashnode&utm_medium=feed)

[Jan 22, 2025](https://hashnode.com/posts/apache-kafka-with-zookeeper-in-django-a-complete-guide/679122ad7eaa0dede96ab3d7)

### [Handling Large Data Sets: The Best Ways to Read, Store and Analyze](https://beyondthebenchmark.hashnode.dev/handling-large-data-sets-the-best-ways-to-read-store-and-analyze?utm_source=hashnode&utm_medium=feed)

Managing Big Data with Dask and Parquet

4 min read · [beyondthebenchmark.hashnode.dev](https://beyondthebenchmark.hashnode.dev/handling-large-data-sets-the-best-ways-to-read-store-and-analyze?utm_source=hashnode&utm_medium=feed)

[Jan 26, 2023](https://hashnode.com/posts/handling-large-data-sets-the-best-ways-to-read-store-and-analyze/63d29e90b6ea63a571c52879)
