Jan 2022 – Sep 2023 · 1 yr 8 mos · Bengaluru, Karnataka, India

• Build and own machine learning models and strategies for risk mitigation in digital payment. Provide thought leadership to democratize and scale ML pipelines and Graph learning across organization.

Sep 2019 – Jan 2022 · 2 yrs 4 mos · Gurgaon, India

• Built and managed a high performing Data Science team at AI Garage. Grew AI Garage to 100+ strength from 20+. I was leading 30 Data Scientists and ML engineers, along with looking after university collaborations and research publications. The AI vision of my team were mainly focussed in the below three directions :
• A. Power MasterCard from within by improving the operational efficiency of different verticals through data driven AI
• B. Build and own AI Powered Risk Assessment solutions for our customer banks
• C. Horizontally grow and scale up research in the areas of Graph Learning, Synthetic Data and ML fairness.
• Apart from these, I was deputed to lead the two streams of research and cyber intelligence product innovation in AI Garage. We harboured our prominence in the AI Research Community by getting 20+ paper published in top tier ML conferences such as IJCNN, IJCAI, ECML, ICDAR, ICLR, KDD, CIKM, ACL, ICMLA etc. We have also won some of the most prestigious data challenges organized by Governments, Universities and Industries in venues like ACM KDD, CIKM, PAKDD, Kaggle Whitehouse COVID prediction and M5 forecasting. We will be organising a workshop on modelling uncertainty in the financial world in ACM CIKM'21.
• Although my team management responsibilities overwhelm me now a days, but to be honest, technology always remains my first love. And fortunately, I am able to remain technically engaged at good depth in quite a few challenging & state of the art ML/DL research problems pertinent to Mastercard around Large Scale Graph Representation Learning, temporal point process, ML Fairness, GAN, Adversarial Attack prevention, RL and multi objective combinatorial optimization. Please connect with me if any of these research areas or product innovation scope interests you and inspires you to sail with us to redefine AI for Payment Network Intelligence.

Apr 2019 – Sep 2019 · 5 mos · Bangalore

• AI driven Network Intelligence : Drive the growth of 5G adoption and reduce huge network maintenance cost enabled by learning from inhouse and telecom operators data.
• Hired and groomed 10+ talents from industry and academia in the short stint.
• Developed a fully functional and deployment ready time series forecasting model for predictive maintenance of Software Defined Network controller. Also worked on to make the model explainable for running root cause analysis.
• One WO patent was filed before I left and recently got granted.

May 2018 – Apr 2019 · 11 mos · Bangalore

• I worked in the business intelligence team to design predictive model for different KPI's of Finance, HRM and CRM. The fine balance between ML model accuracy and explainability was key to the adoption of these models within the organization. And there for the first time in my industrial career, I got the coveted opportunity to be challenged by the non-stationary time series analysis and forecasting in real life : as data shift and concept drift were inherent component of all of these business problems.
• #Automated Dimensional Attribute Ranking System for KPI Metrics
• #Automated KPI Ranking System
• #Forecasting and anomaly detection with multiple dynamic time series data for business analytics application
• #Dynamic thresholding algorithm for non-stationary time series data
• Two patents were filed out of my work in Oracle.

Jan 2018 – May 2018 · 4 mos · Bengaluru Area, India

• Worked on developing business analytics solution for APAC region in the Nextgen team. Built the prototype and POC of an EMI recommender system based on customer spend behaviour and card balance history and present status.

Jan 2016 – Jan 2018 · 2 yrs · Bangalore

• My 1st stint in the industry after finishing my PhD started in the Advanced Technology Lab lead by Dr. Aloknath De. I lead a team of brilliant minds from top IIT's. I mostly worked on building mobile and wearable based healthcare applications, for the users to track and monitor their physical and mental health aspects in daily life. Technically, my research and product innovation portfolio had a perfect mix of signal processing (for denoising sensor data) and Machine learning to model user health profile. Needless to say, the whole ML pipeline had to be real-time. I was fortunate to collaborate with a group of scholars from the diverse fields of Statistics, Computer Vision and Healthcare. We also worked with medical schools for data annotation.
• # Estimation of user alertness level using smartwatch PPG sensor data. Developed a novel signal processing algorithm using wavelet transform to extract heart rate variability form PPG with high reliability and a soft margin multiclass SVM based alertness scoring method.
• # An L1 regularized logistic regression based method for detecting mental stress levels in real time using PPG sensor data with stress level annotated by psychonalysts during different induced mental activities. The stress indicator is already offered as a feature of Samsung Health app in their flagship smartphones. I personally supervised the software engineering team on serving and monitoring the deployed model.
• # Personalized cumulative UV dose estimation system for smartwatch and smartphone devices having the following novelty factors; (a) sensor orientation invariant measurement of UV exposure using a bootstrap resampling technique, (b) estimation of UV exposure using only light intensity (lux) sensor (c) optimal UV exposure dose estimation.
• # Dynamic clustering of high dimensional categorical and sparse data with applications in car analytics and healthcare analytics.
• 3 papers in IEEE EMBC came out of my research in Samsung.

Aug 2011 – Dec 2015 · 4 yrs 4 mos

• Solved a severely ill-posed and under determined inverse problem in Optical Imaging (part of thesis): developed a multi step compressive sensing based optimization algorithm for in-vivo medical imaging using Diffuse Optical Tomography (DOT). The algorithm has been successfully applied for functional brain imaging and prostate cancer imaging. The work got published in Nature. The algorithm provided the highest possible reconstructed image resolution of abnormal tissue growth at its early stage and thereby bring diffuse optical imaging closer to adoption by radiology community for reliable tissue imaging.
• Sparse image reconstruction (part of thesis): established the sufficient condition for sparse image reconstruction in linear inverse scattering problem and analyzed the stability condition in presence of noise.
• Physiological noise removal from brain signal
• Source localization in high-density DOT: detected the sulci region in the somatosensory cortex using time series analysis of the detected light signals at multiple detector sites.
• Functional Brain Connectivity Estimation: estimating information flow in brain from high density EEG signals using directed partial correlation.
• Frequency Domain DOT for Early Prostate Cancer Detection

Aug 2011 – Apr 2015 · 3 yrs 8 mos

• Mathematical Foundation for Electrical Engineering
• Discrete Time System and Digital Signal Processing
• Linear System
• Probability and Random Signals

Jan 2011 – Aug 2011 · 7 mos

• Worked as GRA in Nanofab, UT Arlington for 2 semesters.

Nov 2006 – Aug 2010 · 3 yrs 9 mos · Greater Noida, India

• Performed timing and power dissipation analysis under the best case i.e. high voltage and low temperature and worst case i.e. low voltage and high temperature of the USB, SATA and MIPI transceiver modules developed by ST
• Developed the flow to design timing and power constraint library for ST standard cells consisting of gates and flip-flops of different driving strength
• Optimized the behavioral RTL and post-layout netlist of the digital blocks to avoid set-up and hold violation and at the same time satisfying power and area constraints