May 2020 – Feb 2021 · 9 mos · Bengaluru, Karnataka, India

• Working On HPC performance optimization, benchmarking for HPC application

Sep 2018 – May 2020 · 1 yr 8 mos · Bengaluru, Karnataka, India

• Algorithm development for next-generation Lithography tools.Creating POC and hardware assessment of computing, bandwidth, memory and designed product architecture with pipe-lining of the various module
• Developed 3d printing simulation tool for specific resin emulating and heat dissipation for the specific viscosity and other physical properties in Python. Used clustering methods to model the properties of the various simulations. Uses matplotlib and plotly visualization tools to create an interactive dashboard. Used sklearn ML pipeline to model simulation data.
• Developed Time Series forecasting model to forecast the performance of microlens for performance tuning. Long Short-Term Memory Network (LSTM) a Deep learning RNN was used.
• Death of focus delay and deviation forecasting using Auto-Regressive Integrated Moving Average (ARIMA). Later this is used for simulating the real-time anomaly probability simulation.
• Worked on POC for defect classifier using CNN and tuned the hyperparameter for accuracy. Used sklearn and L2 regularisation for tuning parameters. Create an ML pipeline and used several deep learning algorithms.

Dec 2015 – Sep 2018 · 2 yrs 9 mos · Bengaluru, Karnataka, India

• Responsible for conducting research and developing cutting edge underwriting and portfolio management solutions using High-Performance Computing (GPU and parallel programming) for fluvial and pluvial flood, storm surge, and tsunami risk assessment.
• Building machine learning models for rainfall-runoff simulation using R and CUDA. Build a model for prediction and clustering for flood frequency analysis.
• Proposed product architecture for a project which involves the data management portal for flood data. Improved data mining processes, resulting in a 20% decrease in time needed to infer insights from flood data used to develop analytics model strategies
• Code statistical model for evaluating kernel density estimate using MLCV with Gaussian kernel in R and C language. Parallelized this model in CUDA for efficiency and performance gain.
• Worked on TIME Series regression model using various algorithms like ARIMA, Holt's winter. Implemented LSTM model to predict rainfall-runoff simulation
• Parallelized AffDef model achieves ~5000x speedup

Nov 2013 – Dec 2015 · 2 yrs 1 mo · Bengaluru, Karnataka, India

• Optimizing scientific codes on xeon & xeon phi and CUDA.
• Worked on porting & optimization of Climate Forecast System (v2) over Intel architecture
• Optimization of Genome sequencing application for Intel MIC (xeon phi)

Nov 2013 – Sep 2015 · 1 yr 10 mos · Bengaluru, Karnataka, India

• Optimization and performance improvement of different domains problems into Xeon PHI(MIC) architecture.
• Worked on Molecular Dynamics code, Astrophysics code, Genome sequencing code and some benchmarking code, to improve their performance on Xeon PHI.
• Worked on different profiling and debugging tools some of them are Vtune, ITAC, Intel Inspector XE... etc....

Apr 2013 – Oct 2013 · 6 mos · Chennai

• Worked mainly on openCL implementation and development of complex computational photography project. Got exposure of video and image processing algorithms.
• Responsible for Libra-office calc optimization and porting targeted for AMD HSA arch. Handling team and client interaction. Understanding client requirement and distributing work across engineering team back in china.

Jan 2011 – Dec 2012 · 1 yr 11 mos

• I worked at HPC Links Pvt. Ltd. in the area of High Performance Computing.
• I worked on porting a Weather Modelling and forecasting application CFS to GPU. I also ported the WRF weather forecast model onto Amazon Cloud. I worked primarily with openMP, MPI and CUDA, on Linux and Windows Platforms.
• My responsibilities at HPC Links included hands on programming on CUDA to develop GPU based applications , benchmarking, testing and performance tuning.