Download PDF

Summary

I bring a unique combination of technical and programming depth along with an extensive experience across industries and business functions in AI, optimization and machine learning. I grasp the nuances of a business decision quickly,  am adept at breaking complex business requirements, and implementing actionable algorithms and models.  

I am developing a probabilistic distance clustering algorithm in Python that converges fast and seems to be an alternative to the well-known EM methods in machine learning for de-mixing distributions. 

I am also building LSNLOPT, an extensible and robust framework in C++20 for large scale nonlinear optimization that exploits super-sparsity, uses graph theoretic algorithms for fill-in reduction and addresses common pitfalls.   

I was a hands-on Manager, Decisions Sciences at Kohl’s; where, I used Python and tools like GAMS, GUROBI to build models for assortment planning and improving supply chain across 1100 stores nationwide. My expertise in building complex scalable algorithms helped me to prototype faster and get business approvals for the tradeoffs involved between decision granularity and model complexity. 

I am focused on advancing optimization methods and machine learning algorithms.

Skills

  • Modeling: Machine Learning | Statistical Modeling | Math Programming | Discrete Event Simulation
  • Algorithmic: Online Algorithms | Graph Algorithms | Vehicle Routing | Large Scale Nonlinear Optimization Algorithms
  • Technologies: Python, C++, R, SAS, Java, FORTRAN | GUROBI, CPLEX, COIN-OR | GAMS, AMPL, SIGMA, SLAM | Oracle PL/SQL, Microsoft Transact SQL, Teradata SQL, HDFS – Hive QL.

Work experience

Director - Machine Learning and Optimization

December 2020Present
GS Ravindra, PhD Consultants, Palo Alto, CA
  • Collaborating with UT Austin faculty on building LSNLOPT, an extensible, robust and efficient C++20 framework for large scale super-sparse constrained nonlinear optimization - to shrink algorithm researchers' efforts by 75%.
  • Collaborating with Rutgers faculty in developing probabilistic distance clustering algorithm in python – taking 80% less computational effort than comparable EM methods for de-mixing Gaussian distributions. 
  • Developing innovative randomized algorithms to address graph theoretic problems like fill-in reduction.
  • Built an efficient lookup for matching physician encounter data - over 3.5B rows, the lookups had to performed in a distributed and fault tolerant re-entrant manner.   
  • Implemented health risk scoring for Medicare and Medicaid patients for a Federal Government contractor. To score 8 logit regression models for over 70M patients, devised an efficient algorithm using minimum number of passes.

Manager - Decision Sciences

June 2022May 2023
Kohl’s Corporation, Palo Alto, CA
  • Led the Decision Sciences team in building models and algorithms for merchandizing, marketing, supply chain and logistics.
  • Built a framework for assortment planning - purchase order planning and space optimization for 1100 stores nationwide and six fulfillment centers to improve gross margins.
  • Designed inventory allocation mechanism to stores and fulfillment centers based on near term sales forecast, to reduce markdowns and unmet demand costs. 

Senior Principal Data Scientist, Lead - Machine Learning and AI

October 2018November 2020
Oracle Corporation, Redwood Shores, CA
  • Led the application of AI/ML techniques for User Experience Research, building tools for UX Designers of IaaS, PaaS, and SaaS products.
  • Designed and developed Python tools utilizing libraries such as nltk, spacy, and scikit-learn, which helped conversational UX designers, achieving a "first turn" accuracy of 90% for chatbots of enterprise applications.
  • Executed cloud infrastructure user segmentation and conversion modeling with scikit-learn, successfully identifying the 5.3% of users who upgraded from free trials, aiding in targeted marketing strategies.
  • Reduced time taken by 65% for a 7-member team analyzing click history data of enterprise customers by standardizing path analysis in R and Python.

Principal Data Scientist – Machine Learning and AI

June 2017October 2018
First Tech Credit Union, Mountain View, CA
  • Implemented a next best offer model for marketing in R using random forests and neural networks, achieving a 21% increase in additional product uptake with new customers.
  • Rolled out a delinquency prediction model in R, utilizing random forests and logit, which optimized the 30-member collection team's effort.
  • Modeled probability of default, exposure at default, loss given default in R for a $5B portfolio of mortgages – to assist Chief Risk Officer in computing expected credit loss for federal regulators.

Senior Manager - BI and Data Sciences

February 2014February 2017
Symantec Corporation, Mountain View, CA
  • Developed a comprehensive two-year AI/ML roadmap for a $500M  global customer support operation, targeting a 10% savings in annual costs with innovative solutions for case volume forecasting, agent scheduling, and customer segmentation.
  • Engineered a queueing model to standardize support processes and simplify agent and case states with a potential annual savings of 5-10% that handled long service durations, priority interrupts, rework etc. unaddressed by current vendor tools.
  • Conducted a critical analysis of a board decision to eliminate a premium enterprise service based on gross margins alone. These customers constituted 93% of enterprise segments, had a 12x higher cross sell and 18% higher renewal rates prompting a reversal of the decision.

Director - Data Sciences

November 2012February 2014
Vidya Insights Ltd., Bangalore, India
  • Led the design for an AI/ML startup, focusing on Texas energy retailers such as TXUE and Reliant. Developed meter-level as well as aggregate market demand forecasting models, and proactive customer alert systems for peak demand response. 

Associate Vice President, Practice Leader - Global Advanced Analytics Practice

August 2008October 2012
HCL Technologies, Bangalore, India
  • Invited by the CEO to drive innovation in technology services, successfully reaching Fortune 100 customers with business-focused solutions.
  • Established an advanced analytics practice from the ground up, securing financial services and other US clients, resulting in $6M direct annual revenue and a 59% gross margin, with average margin per person 4.5 times the company- wide average. Built and led a 24-member team, focusing on customer acquisition, cross-selling, segmentation and profiling, market-mix, and text analytics, driving significant business growth and innovation.
  • Managed a 15-member cross-disciplinary team of practice heads to launch a new business process-centric infrastructure management service, including a customer-paid proof of concept for JP Morgan & Chase. Developed innovative components like ProcessWatch and PathFinder process libraries and influenced over $600M in large deals within its first year.

Tenured Associate Professor - Production and Quantitative Methods

August 2004August 2009
Indian Institute of Management, Ahmedabad, India
  • Delivered advanced courses in Mathematical Programming and Statistics to MBA students, and Nonlinear Optimization to PhD students. 
  • Conducted pioneering research on complex graph theoretic problems like the Traveling Salesman Problem, Facility Location and optimal power flow solutions for real-time load dispatching in utility grids.
  • Advised federal agencies and  global logistics companies on complex planning for nationwide operations.

Education

Ph.D. Management Science and Information Systems

University of Texas at Austin - The Red McCombs School of Business, Austin, Texas
  • Research: Large scale discrete and nonlinear constrained optimization.
  • Minors: Distributed Computing, Computational Theory

MBA - Finance and Information Systems

Xavier’s School of Management (XLRI), Jamshedpur, India

BS - Chemical Engineering

Indian Institute of Technology, Varanasi, India

Publications

  • A Fast Tabu Search Implementation for Large Asymmetric Traveling Salesman - OPSEARCH, Springer (2012).
  • Implementing Tabu Search to Exploit Sparsity in ATSP instances. W.P. No. 2008-10-02, Indian Institute of Management, Ahmedabad (2008).
  • HCL Technologies: Employee First, Customer Second - Case Study UVA-OM-1366, Darden Business Publishing, University of Virginia (2008).
  • Scaling Sparse Constrained Nonlinear Problems for Iterative Solvers - W.P. No. 2006-08-06, Indian Institute of Management, Ahmedabad (2006).
  • Scaling Sparse Matrices for Optimization Algorithms - W.P. No. 2006-08-05, Indian Institute of Management, Ahmedabad (2006).
  • Computational experience with a safeguarded barrier algorithm for sparse nonlinear programming - Computational Optimization and Applications, vol 19, pgs 107-120, Kluwer Academic Publishers (2001).
  • A single server queue with cyclically indexed arrivals and service times - Queuing Systems: Theory and Applications, vol 15, pgs165-198, JC Baltzer AG, Science Publishers.
  • INTOPT: an interior point algorithm for large scale nonlinear optimization - Doctoral Dissertation, The University of Texas at Austin.
  • A textbook for managers on using quantitative methods - Mathematical Programming for Managers - TBD.

Patents

  • Detecting wasteful data collection - US 8,825,609 · Issued Sep, (2014).
  • Resource management using environments - US 8,635,624 · Issued Jan, (2014)

About me

  • Personal Interests:
    • Travel, Reading, Hiking, and supporting a Delhi non-profit Niramaya in addressing preventable blindness
  • Fun Facts:
    • I go by initials GS. 
    • I have been serial entrepreneur – once in Austin, TX and twice in Silicon Valley.
    • I have done three trans-continental relocations.
    • I grew up in Kipling country – central India.
    • Along with English, I can speak four Indian languages - Hindi, Telugu, Bengali, and Bhojpuri