MAHESH BUDDE

Accomplished data scientist with AI, machine learning, and software development expertise, leveraging strong analytical skills and experience in diverse projects and internships to drive innovative solutions.

Prediction of Optimized Route for a Ship Traveling in a Seaway using Artificial Intelligence

• Developed and implemented a reinforcement learning model to optimize ship routing, minimizing travel time and enhancing safety by considering dynamic weather and ocean conditions.
• Utilized Deep-Q Network (DQN), Proximal Policy Optimization (PPO), and Deep Deterministic Policy Gradient (DDPG) algorithms, incorporating stochastic weather data and hydrodynamic factors into the simulation environment.
• Achieved significant reductions in travel time and wave heights, demonstrating the effectiveness of AI in dynamic maritime route optimization.

Art Generation with Neural Style Transfer

• Implemented Neural Style Transfer(NST) to generate a new Image from two different Images with the content of one image and with the style of the other.
• Used the Pre-trained VGG-19 Model to recognize a variety of low-level and High-level features
• Separately defined the Content Cost and Style Cost functions and took the relative weightage to Optimize the model

Wordle Solver | AI

• Solving popular five-letter word guessing game, Wordle using Information Theory and entropy.
• Deciding the best word based on information gained and choosing at each level to finally land on the target word.
• Added term frequency weightage to decide to guess the best possible most frequently used word.
• Created a Graphical User Interface to take manual input or choose the best possible next word.

Deep Audio Classifier | Deep Learning

• To identify which areas of forest have the greatest density of capuchin birds using audio clips of bird calls.
• Converted the audio data to waveform and transformed it into a spectrogram using Fourier Transform.
• Visualized the spectrograms and identified the unique signal patterns of capuchin birds.
• Built CNN model on spectrograms to classify whether it's a capuchin bird call (Binary classification).
• Predicted and counted no of capuchin bird calls in a given test audio clip and got an accuracy of 94%.