This project trains a CNN-LSTM model for motion analysis, helping a bot determine the optimal path to reach a goal. Using optical flow (Lucas-Kanade), it captures spatial and temporal motion features. The model ensures smooth trajectory prediction, making it useful for navigation, object tracking, and robotics applications.

Leveraged NVIDIA Isaac Sim and Gazebo simulation environments, alongside PyTorch for reinforcement learning. This initiative focused on training a wheeled robot to autonomously navigate intricate terrains by iteratively developing strategies to evade obstacles and ensure collision-free movement. The project aimed to enhance the robot's operational autonomy and safety in dynamic environments.

Contributed to building a 4-wheeled robot for greenhouse monitoring. Led deep learning model development for pest detection, collaborated on parts design and managed communication protocols to ensure seamless integration and functionality of AI solutions with Raspberry Pi and Arduino platforms.

Developed a local AI chatbot using LLMs (Llama 3 via Ollama) with speech-to-text (STT) and text-to-speech (TTS) for voice interaction. Built with Tkinter, it enables real-time dialogue, emotional detection, and math solving (SymPy), leveraging multithreading for smooth performance.

Built a car detection system using a custom ResNet-50-based model with PyTorch and OpenCV, achieving 96.7% accuracy in diverse images. Focused on robustness across varying lighting conditions, angles, and backgrounds.

Implemented a Q-learning-based reinforcement learning agent in a grid-world environment, optimizing navigation to find the shortest path while avoiding obstacles. Addressed challenges in exploration-exploitation balance and reward sparsity.

Designed and simulated a self-balancing two-wheeled robot using a PID controller and Kalman filter in Simulink. Successfully stabilized the robot after disturbances, demonstrating effective control algorithms.

Developed an IoT-based irrigation system with secure network architecture, automating watering based on temperature and humidity sensors. Simulated in Cisco Packet Tracer, ensuring remote monitoring and cybersecurity.