Sep 2024 – Present · 1 yr 9 mos · San Jose, California, United States · On-site

• Part of AI Data Platform team

Dec 2020 – Aug 2022 · 1 yr 8 mos

• Worked on Capacity forecasting in time series multivariate data & Anomaly Detection in File Access Pattern using Facebook’s Prophet, LSTM models.
• Integrated the pre-trained model in web browser using TensorFlowJs to test inferences.

Jul 2018 – Dec 2020 · 2 yrs 5 mos

• Black Box modeling using Machine Learning for OnTap Sizing
• Implemented end to end modeling pipeline to predict CPU utilized and integrated with
• SPM, a Sizing tool, to improve Sales Margins.
• Prediction is based on real time data collected from ASUP on hourly, weekly granularity which improved the sizing accuracy by 15%.
• Designed & built scalable machine learning lifecycle which involves data procurement, processing, cleansing, feature selection, data modelling, monitoring & performance metrics to ensure continued model stability.
• Implemented feature engineering, model optimization techniques which improves sizing results by 8- 10%.
• Performed A/B testing of various models that drive improvements in model performance.
• Implemented various analytics to gain insights on data and predictability.
• ML algorithms used : Random Forest, CART model, Gradient Boosting(GBM)
• Data Analysis Techniques : Bar charts, Scatter plots, Histograms, Learning Curve, Error Margins plots, Correlation plots
• Tech Stack : NumPy, Scipy, pandas, matplotlib, seaborn, PySpark, SQL, HDFS

Jul 2017 – Jun 2018 · 11 mos

• > Worked as a part of OnTap Performance Sizing Group, simplified SPM, which is used for predicting performance metrics of storage systems for different application workloads, workflow by automating API calls. Hence, reduced team efforts to update currency numbers from 7 days to 1 day.
• > Made Sizing Infrastructure data-driven for performance data and enabled auto generation of access APIs.
• > Implemented ultra cache for performance data to avoid repeated call to db access APIs.