Optimized kNN Inference Engine with KD-Trees and PCA

Just shipped a Retrieval-Augmented Generation (RAG) project and honestly? It's been one of the most interesting things I've worked on. The concept is straightforward: → Feed it documents → Use embeddings for semantic search → Retrieve relevant context from a vector database → Generate answers grounded in your actual data But pulling it all together taught me something important: most "smart" AI systems aren't magic. They're smart pipelines. What I used: • Python • Embeddings for semantic search • ChromaDB (vector database) • A full RAG pipeline This feels like a genuine step toward building real AI applications instead of just experimenting with models. If you've built something similar, I'd love to hear how you approached the trickier parts—especially around retrieval and context quality. GitHub : https://lnkd.in/gwmqvvuT #MachineLearning #AI #LLMs #RAG #Python

Built my first AI-based RAG (Retrieval-Augmented Generation) application 🚀 It’s a simple idea: 📄 Upload a PDF → 💬 Ask questions → 🤖 Get answers grounded in that document While building this, I got hands-on with how RAG actually works behind the scenes: • Extracting text from PDFs (including OCR for scanned files) • Chunking and generating embeddings • Storing vectors in a database for semantic search • Retrieving relevant context and generating answers using an LLM Tech stack: Python, FastAPI, Streamlit, Qdrant, Sentence Transformers, OpenRouter One key learning for me: 👉 The real challenge isn’t just generating answers, but retrieving the right context Still learning and improving, but this was a great first step into building AI systems. 🎥 Sharing a quick demo below 🔗 GitHub: https://lnkd.in/g4nTZu4S Would love your feedback! #AI #RAG #MachineLearning #LLM #Python #BuildInPublic

🏗️ Code isn't just about logic—it's about how you manage data. On Day 2 of the Zenith Edureka #100DaysOfPython challenge, we are tackling the building blocks of every application: Variables and Data Types. As an AI/ML Engineer, I see variables as the "memory" of our models. Whether it’s an integer representing a count or a float representing a neural network's weight, how you define and name your data dictates the quality of your output. Today we deep-dive into: 🔹 Strings & Booleans: Handling text and logical conditions. 🔹 Integers & Floats: The math behind the machine. 🔹 Dynamic Typing: How Python manages memory allocation on the fly. 🔹 Naming Conventions: Why "Snake Case" is the industry standard for professional devs. Mastering these fundamentals is what separates someone who "knows Python" from someone who can "build with Python." Join the Challenge: Watch the tutorial and replicate the code in your VS Code. Drop a comment with "Day 2 Complete" to stay on track. #Day2Of100 #100DaysofCode #PythonForJobs #CodingInterview #PythonBasics #TechCareer2026 #Python #Code

💫 Iris Dataset Multi-Model Classifier An automated machine learning script that evaluates multiple classification algorithms on the classic Iris dataset. This project demonstrates model training, testing, and accuracy comparison using data stored in Excel format. 🚀 Overview This project takes a modular approach to machine learning by iterating through several popular Scikit-Learn classifiers to determine which model performs best on the provided Iris training and testing data. ⚓ Models Evaluated: 1. Random Forest (Ensemble) 2. Decision Tree (Tree-based) 3. Gaussian Naive Bayes (Probabilistic) 4. Extra Trees (Ensemble) 🛠️ Installation & Usage 1. Clone the repository:   https://lnkd.in/dS5Bdvy2 #DataScience #MachineLearning #Python #ScikitLearn #AI #PortfolioProject

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

It’s been a while… but I’m back and still learning 🚀 Today in my AI/ML journey, I explored NumPy, and I’m starting to see why it’s so important. NumPy is a Python library mainly used for working with numbers and arrays (a way of storing multiple values). It makes calculations faster and easier compared to normal Python lists. Some of its uses I came across: - Performing fast mathematical operations - Working with arrays and large datasets - Supporting data analysis and machine learning tasks A simple example: import numpy as np arr = np.array([1, 2, 3, 4]) print(arr * 2) This will multiply all the numbers in the array at once → [2, 4, 6, 8] That’s what makes NumPy powerful—you can do many calculations at once. Still learning… one step at a time. #AI #MachineLearning #NumPy #LearningInPublic #M4ACE #TechJourney

Another step toward efficient AI 🚀 I’m excited to share Inferscale 0.1.1, a lightweight Python package for improving LLM response quality using inference-time scaling. The goal is simple: maximize output quality while minimizing cost. No heavy retraining. No complex setup. Just better results with smarter techniques. If you're working with LLM applications, this could be a valuable addition to your toolkit. Explore it here: https://lnkd.in/giq8KJ5g Feedback is highly appreciated! #ArtificialIntelligence #LLMEngineering #PythonDev #OpenSourceAI #GenAI

Weekly Challenge 12: K-Means Clustering. Finding order in chaos with Unsupervised Learning (K-Means). 12 weeks of coding! To wrap up my first trimester of weekly Python challenges, I dove into Unsupervised Machine Learning. Last week, I coded an algorithm that learned from labeled data (KNN). But what happens when you have thousands of data points and zero labels? You let the math find the patterns. I built the **K-Means Clustering** algorithm from scratch using only `NumPy` and Math. The Optimization Loop: 1 Drop 'K' random centroids onto a dataset. 2 Assign every data point to its closest centroid. 3 Calculate the mean (average) of those points, and move the centroid to that new center. 4 Repeat until the centroids stop moving (Convergence!). Instead of just plotting the final result, I programmed Matplotlib to draw the trajectory of the centroids. In the graph below, you can see the exact path the AI took (the dotted lines) to find the perfect center of each cluster! Consider this scenario: We want to understand our clients' true preferences. We test the market by offering 3 initial color options and observing their reaction. Based on that feedback, we dynamically adjust our offerings to align perfectly with what the clients actually want Full source code on my GitHub:https://lnkd.in/gYs2adtN #MachineLearning #Python #DataScience #ArtificialIntelligence #KMeans #UnsupervisedLearning #Algorithms #CodingChallenge #Optimization

✨Day 3 of my Pattern-wise DSA Journey 🔢 Problem Name: Maximum Sum Subarray of Size K 🧩 Pattern Used: Sliding Window Technique 🐢 Brute Force Idea: Generate all possible subarrays of size K and calculate their sums one by one. This works but takes O(n·k) time, which is inefficient for large inputs. ⚡ Optimal Idea: Used a sliding window of size K. Instead of recalculating the sum every time, subtract the element going out of the window and add the new element coming in. This reduces time complexity to O(n). 🧠 What I Learned: • How sliding window avoids repeated calculations • Turning nested loops into a single pass • Importance of window size in optimization • Writing efficient and scalable solutions 💡 Don’t recompute—reuse. That’s the power of sliding window. 🔥 Building consistency, one pattern at a time. #DSA #CodingJourney #ProblemSolving #Python #AI #Learning #Consistency #SlidingWindow

From complex policy documents ➡️ to intelligent insights with AI 🤖📄 I’m excited to share my project “PolicyLens – AI Policy Adaptation Studio”, developed using Python and VS Code. This system is designed to simplify and transform policy documents into meaningful, scenario-based outputs. It helps users quickly understand policies and generate actionable insights based on real-world situations. 💡 Key Features: • Upload and analyze policy documents (PDF/TXT) • Automatically generate structured summaries • Scenario-based policy generation • Customizable inputs (target audience, constraints, focus areas) • Clean and interactive user interface 🚀 Through this project, I improved my skills in Python development, AI-based text processing, and building practical solutions for real-world problems. 📽️ Here’s a quick demo of the system in action Feedback and suggestions are always welcome 🙌 #Python #AI #MachineLearning #DataScience #Innovation #StudentProject #Tech #LearningJourney