- developed translator for crossgl language to translate the code from metal, hlsl, opengl shader to crossgl universal shader.

- Collaborating on test case fixes and critical tasks with the Ivy team, involving deep dives into transpiler and compiler code, and reviewing pull requests.- developed a chatbot to compare and rank the top LLM with a global leaderboard user can ask query to two llm same time can vote the winner- contributed in handling the complex dtype at tensorflow , jax and paddlepaddle backend imroved the some SHLO in xla for complex dtyep.

- open source a machine learning tool that helps optimize models, converts code between frameworks, and allows writing code independent of specific ML frameworks.- Contributed with across JAX, NumPy, PaddlePaddle, TensorFlow, PyTorch, and Ivy function API, enhancing machine learning frontend APIs- Implemented a range of non-linear activations in Ivy's API (e.g., HardTanh, Softshrink) for versatile neural network customization. Contributed to Paddle's frontend with functions like Reciprocal, Tanh, and Remainder. Enhanced JAX by integrating Std, Var, ELU Activation for statistical analysis and varied activations.

- Developed precise Rhino height estimation models and a robust 98% accurate detection pipeline.- Led model development, optimized depth estimation for accurate height predictions, and designed the Rhino identification pipeline.- Achieved <2% error in Rhino height estimation, 98% accurate Rhino detection, and utilized depth estimation for improved accuracy.

- Created and deployed an interactive Tableau dashboard presenting COVID-19 death and survival- Designed intuitive interface, improving data exploration by 40%. Implemented robust email filtering achieving 98% precision, 95% recall, F1 score of 0.96 for enhanced user experience.- Interactive COVID-19 Dashboard: Enabled exploration of regional COVID-19 trends via an intuitive Tableau interface. Accurate Spam Email Classifier: Leveraged XGBoost and SVM algorithms for 98% precision, reducing spam by 60%. Enhanced Email Filtering: Achieved 95% recall and an F1 score of 0.96, significantly improving email filtering accuracy and user satisfaction.