Projects
AI Agents, Generative AI applications, Machine Learning projects, Deep Learning systems, and AI engineering experiments.
Computer-Use-Browser-Automation-Agent
Computer-use agent for real-world browser automation. It can navigate websites, click elements, type text, fill forms, interact with pages, and complete multi-step tasks autonomously. Built to explore practical agentic AI systems that can perceive, reason, and act inside real web environments.
Learn more →Autonomous-Coding-Agent
An autonomous coding agent that takes GitHub issues, understands repository context, plans fixes, edits code in a sandbox, runs tests, self-corrects failures, and automatically creates branches, commits, and pull requests with generated summaries and test results.
Learn more →Self-Improving-Agent-with-Automated-Prompt-Promotion
A production-oriented AI agent that automatically evaluates multiple prompt variants against a regression suite, compares their performance, detects improvements, and promotes the best-performing prompt. It demonstrates automated prompt optimization, evaluation-driven development, regression protection, and continuous agent improvement.
Learn more →Cross-Service-Observability-Evaluation-Platform
A production-grade platform for tracing AI agent workflows across services, correlating requests end-to-end, collecting structured telemetry, and evaluating reliability, latency, cost, failures, and tool usage. Built to make distributed AI systems measurable, debuggable, and continuously improvable.
Learn more →Distributed-Event-Driven-Agent-Platform
A production-oriented distributed event-driven agent platform where AI agents react to events through a Kafka-compatible message bus. Built with Redpanda, Python, and event-driven architecture to enable scalable, decoupled, fault-tolerant agent workflows.
Learn more →Persistent-Multi-Tenant-Memory-Service
Production-ready multi-tenant memory microservice for AI agents with semantic & episodic memory, vector search, ranked retrieval, automated compression, and reusable SDK.
Learn more →Enterprise-Knowledge-Graph-Agentic-RAG
Production-grade Enterprise Knowledge Graph Agentic RAG system featuring Hybrid Search, Neo4j Knowledge Graphs, Qdrant Vector Search, Multi-step Retrieval, LLM-powered reasoning, Evaluation Harness, Langfuse Observability, and Docker deployment.
Learn more →Movie-Recomender-System-TFIDF-Consine-Similarity-Streamlit
The application provides personalized movie recommendations using a content-based filtering approach. By loading a pre-trained TF-IDF matrix and movie dataset from a pickle file, it calculates cosine similarity between plots. Features include URL manipulation, a search interface, and discovery landing page.
Learn more →Handwritten-Digit-Recognition-SVM-Tkinter-
A Python desktop app that uses SVM (Machine Learning) to recognize handwritten digits. Draw on the Tkinter canvas, and the model predicts the number in real-time. Features: Smooth drawing UI Scikit-learn SVM model MNIST-style image processing Instant prediction Tech: Python, Tkinter, Scikit-learn, PIL.
Learn more →mcp-native-tool-platform
A production-ready MCP (Model Context Protocol) platform for building AI agents with dynamically discoverable tools, prompts, resilience, caching, and Dockerized MCP servers.
Learn more →Real-Time-Emotion-Detection-ResNet50-CBAM
A real-time facial emotion recognition system built using PyTorch, ResNet50, and CBAM attention mechanism. The project uses MediaPipe for fast face detection and OpenCV/Streamlit for real-time inference and visualization. The model detects multiple human emotions from images or live webcam feed with high accuracy and efficient performance.
Learn more →ai-surveillance-system-using-yolo-bytetrack
AI-powered surveillance system built using YOLO for real-time object detection, DeepSORT for multi-object tracking, and custom rule-based logic for security event analysis. Detects intrusion, loitering, unauthorized access, and suspicious activities from CCTV feeds, providing real-time alerts and actionable insights for enhanced security monitoring
Learn more →ReAct-Agent-v1
Production-grade ReAct AI Agent with tool calling, web search, FastAPI backend, Docker deployment, Langfuse observability, and evaluation harness.
Learn more →Crop-Recommendation-System-XGboost-classifier-Streamlet-
This Streamlit-based application is a smart **Crop Recommendation System** that leverages an XGBoost machine learning model to help farmers optimize their yield. By inputting key soil and climate data—such as **Nitrogen, Phosphorus, Potassium, Temperature, and pH**—the app provides an instant, data-driven suggestion for the best crop to plant.
Learn more →