![](images/noisy_env.jpg)
Multiple Speaker Recognition in noisy environment
Classified 47 speakers in a noisy environment based on the Interspeech 2019 VOiCES Challenge dataset using Random Forests in Pyspark with mean per class error of 0.196.
Software/ML Engineer
A Senior Data Engineer at LeanTaaS and seasoned Data Scientist, I leverage my expertise in business analysis, cloud backend engineering, and machine learning to drive innovation and efficiency. Having had a successful career with previous roles at Pure Storage and Checkpoint Software Technologies, I've utilized my skills to reduce costs, streamline processes, and enhance cybersecurity, respectively.
In my current role at LeanTaaS, I've designed and implemented a real-time data pipeline for hospital bed capacity management. My leadership helped maintain operational efficiency and facilitated significant expansion of the company. My past experiences at Honeywell Technology Solutions as a Tech Lead and Pure Storage as a Data Analyst allowed me to develop strong leadership skills and significant cost-saving initiatives, shaping me into a proficient problem-solver and strategist.
Academically, I hold a Masters in Data Science from the University of San Francisco and a B.Tech. in Electrical Engineering from IIT Hyderabad. My technical proficiencies range from Python, SQL, and JavaScript to Python Libraries like Scikit-Learn, NumPy, Pandas, PyTorch, and others. In the realm of cloud computing, I'm well-versed with AWS, GCP, Airflow, Git, Jenkins, and GraphQL. I'm excited about the endless possibilities that the interplay of data science and cloud computing holds for the future.
Classified 47 speakers in a noisy environment based on the Interspeech 2019 VOiCES Challenge dataset using Random Forests in Pyspark with mean per class error of 0.196.
Automated the task of cataract patient eye diagnosis by segmenting blood vessels in the choroid layer of the eye in MATLAB. Paper published in European Journal of Ophthalmology
Trained a multi class support vector machine to recognise six facial expressions happy, surprise, sad, angry, disgust, fear by extracting features around the mouth and eyes regions of face. Used a pyramid of local binary patterns on the low resolution images of Cohn Kanade+ dataset achieving 70% accuracy(highest) for happy and sad expressions.
Placed 6th in an in-class Kaggle competition to predict heart rate and pressure in critically ill patients using XGBoost and Neural Networks with an R-squared value of 0.928.
Using Pyspark and Plotly, analysed and visualised School Donations Project data that matches classroom charity projects to the most probable donors nationwide.