Muhammad Aqeel

26/12/2025

Case Study: Automated YouTube Video Summarization using n8n



Overview

This project focuses on building an automated n8n workflow that converts YouTube videos into clear, structured summaries. The solution allows users to submit a YouTube video URL through a simple form and automatically generates a summarized version of the video transcript, which is then saved to Google Docs for easy access and review.



Problem Statement

Manually watching long YouTube videos and extracting key points is time-consuming. Users needed an automated solution to quickly summarize video content and store it in a readable format without manual effort.



Solution

An end-to-end automation workflow was created using n8n that:

Accepts a YouTube video URL via a form

Fetches the video transcript using an external API

Converts the transcript into a single clean paragraph

Uses an AI model to generate a concise, structured summary

Automatically saves the summary to a Google Docs document



Workflow Architecture

1: Form Trigger : Collects the YouTube video URL from the user

2: HTTP Request (Apify API) : Retrieves the full video transcript

3: JavaScript Code Node : Merges transcript text into one paragraph

4: OpenAI Model : Summarizes the transcript into key points and takeaways

5: Google Docs Node : Inserts the final summary into a shared document



Tools & Technologies

n8n (Workflow Automation)

Apify YouTube Transcript Scraper API

OpenAI (GPT-3.5 Turbo)

Google Docs API

JavaScript (Data transformation)

Key Features

Fully automated, no manual intervention

Clean and human-readable summaries

Supports long-form video content

Centralized documentation in Google Docs

Scalable for multiple videos

Results & Impact

Reduced video review time by 80%+

Improved productivity for content research

Enabled fast knowledge extraction from video content

Created a reusable workflow for future summarization needs

06/10/2025

Today I am hands-on practicing AI Agents! 🤖

I built three intelligent agents that can collaborate and solve different types of tasks automatically using LangGraph and LangChain frameworks.

🧠 Agents I created:
1️⃣ Research Agent — Uses DuckDuckGoSearchRun to fetch real-time information for general or current queries.
2️⃣ Add Agent — Performs addition of given numerical values.
3️⃣ Multiply Agent — Handles multiplication-related tasks.

✨ When a user asks a question:

If it’s a general question → Research Agent takes charge.

If it’s about addition → Add Agent executes.

If it’s about multiplication → Multiply Agent responds.

The agents are managed by a Supervisor Agent, which decides which expert should handle the query based on its type — a perfect demonstration of multi-agent collaboration in action! 🚀

03/10/2025

03/10/2025

I am practicing AI Agent in LangChain to enhance my skills in building intelligent systems.

For this, I worked with multiple tools including:

🔍 DuckDuckGoSearchRun – for real-time search results
📖 WikipediaAPIWrapper – for structured knowledge
💾 SQLite3 – for database interaction
📚 ArxivQueryRun – for accessing latest research papers
This practice is helping me understand how AI Agents can combine reasoning with external tools to create more powerful and context-aware applications.

22/09/2025

Project Name
🔍 Deepfake Detection using Transfer Learning (ResNet50)

Overview
In this project, I developed a deep learning model capable of detecting real vs. fake images. With the rise of AI-generated media, detecting manipulated or deepfake content is crucial for ensuring authenticity in digital platforms. This project leverages computer vision and deep learning techniques to classify images into real or fake with high accuracy.

Features
✅ Preprocessing of image frames using OpenCV
✅ Data handling and splitting with Pandas, NumPy, and Scikit-learn
✅ Model training with PyTorch (ResNet50 backbone)
✅ Training visualization using Matplotlib & Seaborn
✅ Model interpretability with torchsummary & torchview
✅ GPU support for faster training

Model Training
Base Model: ResNet50 (pretrained on ImageNet)
Approach: Transfer Learning
Custom Layers:
Fully Connected Layers (1024 → 512 → 1)
Batch Normalization & Dropout for regularization
LeakyReLU activations for better feature learning
Loss Function: Binary Cross-Entropy
Optimizer: Adam / SGD
Output: Sigmoid activation for binary classification

Results & Visualization
📊 Training and validation accuracy were tracked across epochs.
🖼️ Below are sample images from the dataset (real vs. fake) and model prediction visualizations.

07/06/2025

Eid Mubarak to All Muslim

08/05/2025

🔥 Project Title
Superhero Butler AI – SmolAgents Party Planner for Wayne Mansion

🧠 Project Overview
Ever wondered how Alfred might use AI to plan a party at Wayne Manor?
I built an intelligent assistant using smolagents that acts like a digital butler – handling everything from music recommendations to menu planning for themed parties!

🛠️ Description
This AI agent, powered by Hugging Face models and custom tools, uses internet search, menu planning, theme suggestion, and even mock catering services to orchestrate events.
It's an AI-driven event planner for the most elite gatherings in Gotham.

🌟 Features
🕵️ Internet search for top party music using DuckDuckGoSearchTool()

🍽️ Menu suggestions for casual, formal, and superhero-themed parties

🛎️ Custom catering recommendation logic (e.g., "Wayne Manor Catering")

🦸 Superhero-themed party idea generator with Tool subclassing

📄 Conversational interface powered by CodeAgent with multi-step reasoning

🧾 Conclusion
This was a fun way to combine tools, custom logic, and language models to simulate a real-world use case.
Next step: turning it into a voice assistant for immersive event planning!
🎉 Gotham’s parties just got smarter.

Sir Irfan Malik , Sir Dr. Sheraz Naseer , Sir Muhammad Haris

08/05/2025

🔥 Project Title
Superhero Butler AI – SmolAgents Party Planner for Wayne Mansion

🧠 Project Overview
Ever wondered how Alfred might use AI to plan a party at Wayne Manor?
I built an intelligent assistant using smolagents that acts like a digital butler – handling everything from music recommendations to menu planning for themed parties!

🛠️ Description
This AI agent, powered by Hugging Face models and custom tools, uses internet search, menu planning, theme suggestion, and even mock catering services to orchestrate events.
It's an AI-driven event planner for the most elite gatherings in Gotham.

🌟 Features
🕵️ Internet search for top party music using DuckDuckGoSearchTool()

🍽️ Menu suggestions for casual, formal, and superhero-themed parties

🛎️ Custom catering recommendation logic (e.g., "Wayne Manor Catering")

🦸 Superhero-themed party idea generator with Tool subclassing

📄 Conversational interface powered by CodeAgent with multi-step reasoning

🧾 Conclusion
This was a fun way to combine tools, custom logic, and language models to simulate a real-world use case.
Next step: turning it into a voice assistant for immersive event planning!
🎉 Gotham’s parties just got smarter.

Sir Irfan Malik, Sir Dr. Sheraz Naseer , Sir Muhammad Haris

06/05/2025

🧵 Tumor Detection using YOLOv11 & SAM – End-to-End Deep Learning Project

1️⃣ Project Title:
🔬 Automated Brain Tumor Detection using YOLOv11 & SAM for Medical Imaging

2️⃣ Project Work:
🚀 Trained a custom YOLOv11 model (yolo11n.pt) on a brain tumor dataset (glioma, meningioma, pituitary tumors).
📍 Used a detailed data.yaml config to define the dataset and classes.
🧠 Enhanced YOLO's object detection results by integrating Meta AI's SAM (Segment Anything Model) for precise segmentation.

3️⃣ Key Features:
✅ YOLOv8 for fast, real-time object detection.
✅ SAM for fine-grained segmentation on YOLO-predicted bounding boxes.
📸 Image-level predictions and masks are saved and visualized.
⚙️ GPU-accelerated training and inference for high performance.

4️⃣ Conclusion:
💡 This fusion of detection (YOLOv8) and segmentation (SAM) shows great potential for clinical decision support tools.

17/04/2025

🎉 Exciting News — Course Completed! 🎉

I’m happy to share that I’ve successfully completed the "AI for Python Beginners" course! 🚀

Throughout this journey, I learned the core concepts of Artificial Intelligence and Python programming, which really helped strengthen my foundations.

Aqeel Islam, congratulations on completing AI Python for Beginners!

16/04/2025

🚀 Introducing Rent Calculator: Simplify Expense Splitting for Shared Living! 🏠💸

📌 App Name: Rent Calculator
🔍 Overview:
Living with roommates? Splitting rent and utility bills can get tricky. Rent Calculator makes it effortless to divide monthly expenses fairly — no more confusion or manual math!

📝 Description:
Built using Flask, along with HTML, CSS, and JavaScript, this app takes essential house expenses (rent, food, electricity, gas, water) and divides the total among all housemates. Whether you're living in a PG, flat, or shared house — this tool is made for you!

✨ Features:
✅ Expense Input Form – Add rent, utility bills, and number of people
✅ Per Person Share Calculation – Fair, fast, and accurate
✅ Responsive UI – Clean design with mobile-friendly interface
✅ Error Handling – Prevents invalid inputs for a smooth experience
✅ Easily Extensible – Future-ready: can add charts, login, or download options

🔥 Conclusion:
The Rent Calculator brings clarity to shared living costs. It's quick, reliable, and made with real-life needs in mind.