Harshil Gupta

Building Scalable Solutions, Driving Innovation

Developing AI Agents @Salesforce

[Harshil Gupta @HarshilG28]

Act: Execute trades via Zerodha API

Learn: Improve strategies based on outcomes

The agent becomes a autonomous trading partner, not just an analysis tool. This is where AI + Finance gets really powerful.

#HarshilLearnsAI #AgenticAI #AI #FinTech #LLM

[Harshil Gupta @HarshilG28]

Day 4 of #HarshilLearnsAI

Agentic AI is the next evolution beyond what I've been building 🚀

The core difference:

Traditional AI: Input → Output

Agentic AI: Goal → Perceive → Reason → Act → Learn

#HarshilLearnsAI #LLM #AgenticAI #Fintech

[Harshil Gupta @HarshilG28]

Instead of my current setup where I ask Claude to analyze portfolio data, an agentic system could:

Perceive: Monitor market conditions 24/7
Reason: Analyze trends using my risk parameters

#HarshilLearnsAI #AgenticAI #AI #FinTech #LLM

[Harshil Gupta @HarshilG28]

🔗 API Glue - The Zerodha Kite integration is all REST, but wrapping it with smart LLM context makes it conversational.

I could ask for today's market trends to Claude and place an order via Kite Python SDK.

#HarshilLearnsAI #AI #Claude #MCP #zerodha #Kite

[Harshil Gupta @HarshilG28]

Day 3 of #HarshilLearnsAI
Unpacking the black box. Today was all about peeling back the layers on how Claude + MCP + Zerodha actually talk to each other under the hood.

Here’s what I focused on:

[Harshil Gupta @HarshilG28]

🧠 Stateful LLMs – Dug into how session continuity works when I was using Claude as a portfolio co-pilot. It remembers just enough without overstepping. It was not an average stateless chatbot. It remembered what I covered previously.

[Harshil Gupta @HarshilG28]

This blend of LLMs + APIs is 🔥
Amazing 🙂

Tomorrow, I plan to:
Go through the documentation to understand more about how it all works under the hokd

#AI #Python #FinTech #Zerodha #Claude #HarshilLearnsAI

[Harshil Gupta @HarshilG28]

🚀 Day 2 of #HarshilLearnsAI
Diving deeper into AI x Finance. Today’s focus: building a smarter portfolio analysis tool using LLMs and brokerage APIs.

Let’s talk about MCP, Claude, and Zerodha. 👇

[Harshil Gupta @HarshilG28]

👨‍💻 On the backend, I started exploring the Kite Connect Python SDK.
Wrote basic scripts to:

Get holdings
Pull live market data
Prep for trade triggers down the line

Hands dirty. Momentum building.

[Harshil Gupta @HarshilG28]

Excited to dive deeper into LLM internals and agent architectures next!
Goal for 1st week: Build foundational understanding of LLMs + agents. Get hands dirty with tools.
#LearnInPublic #LLM #BuildInPublic #HarshilLearnsAI

[Harshil Gupta @HarshilG28]

Day 1 of my #AI learning sprint: Watched Andrej Karpathy’s “Intro to Large Language Models”.

Key takeaways:

LLMs like ChatGPT, Claude, and Bard are powered by massive neural networks trained on huge internet datasets.

[Harshil Gupta @HarshilG28]

LLMs “dream” new content by predicting the next word, sometimes hallucinating plausible but incorrect facts.

Security risks: prompt injection, jailbreaks, and data poisoning are real challenges for LLM deployment.

[Harshil Gupta @HarshilG28]

Excited to start my #AI learning journey! 🚀 I will be sharing daily updates, resources, and project progress as I dive into topics like LLMs, embeddings, and building AI agents. Follow along and let’s learn together! #LearnInPublic #100DaysOfAI

[Nauseam (in sf!) @ChadNauseam]

"A calculator app? Anyone could make that."

Not true.

A calculator should show you the result of the mathematical expression you entered. That's much, much harder than it sounds.

What I'm about to tell you is the greatest calculator app development story ever told.

[Nick Bilton @nickbilton]

They need to create a Pulitzer category for headlines just for whoever wrote this.

[Harshil Gupta @HarshilG28]

@akshaymarch7 https://t.co/7nd8xDGZFl