Sanjay Kumar Katakam’s Post

The Art of Focus: Mastering Image Cropping with NumPy! 🎯✂️ Day 86/100 In a world of data noise, the ability to focus on what matters is a superpower. For Day 86 of my #100DaysOfCode journey, I explored Region of Interest (ROI) Extraction. In Computer Vision, we don't always need the full picture. By using NumPy Array Slicing, I can 'zoom in' on specific coordinates to isolate faces, text, or objects for further analysis. Technical Highlights: 🎯 ROI Identification: Mastering the coordinate system to pinpoint and extract sub-matrices from large image arrays. ✂️ Precision Slicing: Leveraging Python's [start:stop] syntax to perform lossless cropping in microseconds. ⚡ Computational Optimization: Learning why reducing image size via cropping is the first step in high-speed object detection. 🤖 AI Preprocessing: Understanding how cropping helps prepare datasets for deep learning models by removing irrelevant background noise. Do check my GitHub repository here : https://lnkd.in/d9Yi9ZsC #100DaysOfCode #ComputerVision #NumPy #Python #BTech #IILM #AIML #ImageProcessing #DataScience #SoftwareEngineering #LearningInPublic #WomenInTech

Using python scripts to parse YT transcripts so my Positioning reviews and teardown videos can live on my blog properly. Using tech in a meaningful way to give you an edge in business is pretty critical. All of my current clients, we're bolting on AI consulting sessions in addition to wiring in Positioning and creating an unfair competitive edge. It just makes sense. V1 Blog live. https://lnkd.in/eqHVpK-g

A House Price Prediction Model using Linear Regression! I recently put together a complete, end-to-end Google Colab notebook demonstrating how to predict real estate prices using machine learning. It's fascinating to see how math and data come together to model real-world markets. Data Preprocessing: Handling missing values, feature scaling, and encoding. Model Building: Training a Linear Regression model from scratch using Scikit-Learn. Evaluation: Analyzing the model's performance using metrics like RMSE and R-Squared to ensure accuracy. The biggest takeaway for me was realizing how heavily feature engineering impacts the final model's accuracy. Grateful to IncodeVision for giving me a project that pushed me beyond just writing code! #MachineLearning #DataScience #LinearRegression #Python #GoogleColab #RealEstateTech #IncodeVision

Professional & Technical (Best for showcasing skills) Headline: Automating Image Recognition with TensorFlow 🌸 I recently worked on a flower classification project using the tf_flowers dataset. This project allowed me to dive deep into the mechanics of computer vision and efficient data handling. Key highlights of the implementation: Data Pipeline: Utilized tensorflow_datasets for a seamless 80/20 train-test split. Preprocessing: Implemented a custom preprocess function to normalize pixel values and resize images to $150 \times 150$, ensuring consistency across the neural network. Prediction & Validation: Built a visualization loop using Matplotlib to compare the model’s predicted labels against actual data points. It’s exciting to see how a few lines of Python can transform raw pixel data into accurate classifications. Looking forward to refining this further with transfer learning! #TensorFlow #DeepLearning #ComputerVision #Python #DataScience #MachineLearning

Just completed a House Price Prediction project using Machine Learning 🏠📊 I built an end-to-end pipeline using Linear Regression to predict housing prices, focusing on clean preprocessing and feature engineering. 🔹 Key highlights: - Data cleaning & outlier removal - Feature engineering (house age, room ratios) - Categorical encoding using OneHotEncoder - Model training with Scikit-learn - Evaluation using R² Score (0.70) and RMSE (~149K) This project helped me better understand how preprocessing and feature engineering directly impact model performance. 📂 Check out the project on GitHub: https://lnkd.in/dJHP8X9h #MachineLearning #AI #DataScience #Python #ScikitLearn

"Unknown" ≠ a category. It's a mask 🤦♀️ . In EDA, filling missing categorical values with "Unknown" is a common trick, it keeps groupby() clean and visualizations readable. But before imputation or modeling: df[col] = df[col].replace("Unknown", np.nan) Why? 🤔 Because the model treats "Unknown" as a real category and learns spurious patterns from it. The distinction that matters: "Unknown" → display solution → EDA only NaN → statistical truth → imputation & ML 👉 Always restore the truth before modeling. #DataScience #MachineLearning #DataCleaning #DataQuality #Python

Day 3 of building AI agents on 8GB RAM. Today: My Meeting Summarizer now reads files. Drop a .txt, .docx, or .csv → AI extracts decisions, action items, next meeting. No copy-paste. No formatting. Just drag and drop. Built with: - Python - Ollama (local AI) - Qwen 1.5B - 50 lines of code Cost: $0 Runs on: My 8GB laptop Tomorrow: Adding web search so the AI can research topics mentioned in meetings. #AI #BuildInPublic #LocalAI #Python #Automation

I’m excited to share the latest version of NTQR. A Python package for the logic of unsupervised evaluation of classifiers. Want to get started in under 60 seconds? Here is the "Quick Start" flow to get the documentation and interactive notebooks running locally: 1️⃣ Install the package: pip install ntqr 2️⃣ Set up your workspace: Navigate to the folder where you want your tutorial notebooks to live: cd path/to/your/working/directory 3️⃣ Fetch the docs: Run the built-in helper to copy all tutorial notebooks into your current folder: ntqr-docs 4️⃣ Launch & Learn: Open the environment and dive into the examples: jupyter notebook You can see the notebooks at readthedocs.io (NTQR doc page: https://lnkd.in/eugreNDd). The notebooks walk you through the math and the code, making it easy to apply these techniques to your own AI evaluations of classifiers. Give it a spin and let me know what you think! 👇 #Python #DataScience #MachineLearning #AI #OpenSource #NTQR #FormalVerification #AIEvaluation #UnsupervisedEvaluation

Just finished Anthropic’s Introduction to Model Context Protocol — definitely worth the time. Learned how MCP lets AI models like Claude connect with tools and data without messy custom integrations. Implementing the three core building blocks — tools, resources, and prompts — using Python was a great hands-on experience. It’s free on Anthropic’s learning portal. If you’re into building smarter AI workflows, it’s a great place to start. #MCP #Anthropic #Python #AI #LLM #DeveloperTools #ContinuousLearning

6 months ago I had no idea what a "state graph" was. Today I just finished writing complete LangGraph notes — from zero to production-ready agents. The thing nobody tells you about LangGraph: It's not about AI. It's about state machines. Once that clicks, everything else is easy. My notes cover the mental models, not just the syntax. Because syntax you can Google. Thinking like an engineer — that takes time. Sharing the full PDF with anyone who needs it. What topic should I cover next? 👇 #LangGraph #AIEngineering #Python #LearningInPublic #AIAgents #SoftwareEngineering