Dec 2023 – Aug 2025 · 1 yr 8 mos · Mumbai, Maharashtra, India · On-site

• Startup Leadership: Spearheading the AI vision for disruptive AI-factory agents.
• Leading product design, backend engg., noSQL DBs, workflow orchestration, DevOps, cloud optimization, automation, testing. Leveraging Gen AI to drive seamless, multi-marketplace, 360-degree e-commerce. Enabling companywide AI-driven decision-making across strategy, marketing, sales, operations, customer success.
• 1) Gen AI Agents (Listed under MIT Nanda – Internet of AI Agents):
• Spearheading the vision of a dynamic AI Factory Ecosystem — a network of self-learning, autonomous agents orchestrating intelligent decision-making and optimization across e-commerce dimensions, internal CI/CD pipelines, user experience, etc.
• 2) Generative AI - One-click Content Generation: Developed a cutting-edge system, empowering the smallest of sellers with just a cell phone, to convert multilingual audio and image inputs into fully optimized product listings; across titles, descriptions, lifestyle/infographic images, multimodal content, videos and attributes - with one-click AI marketplace integrations.
• 3) Content Grading & Enhancement: Pioneered a predictive modeling-based grading system that evaluates novel critical dimensions for text as well as images; and provides one-click listing enhancements.
• 4) What if Scenario AI Dashboards: Developing advanced Predictive Modeling Driven dashboards for interpretability and recommendations for sales maximization
• 5) Voice Based Query Engine: Developed a Generative-AI voice engine that interprets local language audio queries, generates executable code, and delivers visual insights and output audio in the input language.
• 6) Generative AI-Based Competition Mapping and Alerts: Enabling Real-Time Insights for sellers to stay ahead of competitors.
• 7) Sentiment Analysis: Developed a Generative AI-based system to provide nuanced insights and actionable multi-dimension recommendations to outgrow competition.

Nov 2022 – Nov 2023 · 1 yr · Mumbai Metropolitan Region

• Startup Leadership: Set up (0 to 1) the end-to-end omni-dimensional vision for data science, machine learning and Generative AI, for a cross-border e-commerce company with 300+ brands. Led the full product development including the Frontend, Backend APIs, Data Layer Setup and Data Science.
• 1) Data Layer Setup: Designed and implemented large-scale, real-time data retrieval systems from the ground up, establishing the foundational data layer for the e-commerce platform.
• 2) Automation & Analytics Tools: Developed a suite of automation-driven analytics tools to empower internal stakeholders with actionable insights for strategic decision-making.
• 3) AI-Driven Strategic Insights Tool: Led the development and deployment of a novel DS and Generative AI driven Strategic Insights Tool (along with a chatbot) for Indian e-commerce sellers (aiming to expand globally), providing comprehensive insights across countries, marketplaces, imports and exports, cross-country category mappings, best seller benchmarking, pricing, revenues, HSN classification and compliances.
• 4) AI-powered Chatbot: Developed a chatbot to help answer unstructured natural language seller queries, with features such as next best questions and persistent memory
• 5) Predictive Modeling Driven Strategic Product Optimization: Developed Novel strategic Insights for revenue prediction for cross-country marketplace product launches and unravelled the revenue contribution of product attributes to enable product attribute optimization
• 6) ML Framework for Seller Journeys: Developed a framework for ML based solutioning for the gamut of use cases across the lifecycle journey of the Indian sellers launched internationally
• 7) ML Control Tower: Envisioned and designed the analytical framework for a ML Control Tower Ecosystem for continuous re-optimitization of the seller ecosystem
• 8) AI Driven Computer Vision: Envisioned and Initiated the development of novel computer vision solutions for e-commerce leveraging Generative AI

Nov 2020 – Nov 2022 · 2 yrs · Mumbai, Maharashtra, India

• Company Revenue: 2+ Billion USD.
• Led ground-up innovation for product as well as service teams by developing advanced novel applied AI (on Big Data) for a multitude of industry verticals - including technology, e-commerce, travel, hospitality, investment management, insurance and agriculture - with Multi-Million USD impacts.
• 1) Journey Optimization: Atlas (SaaS Product) - Envisioned and Developed Novel Advanced-ML solutions (as well as interpretable visualizations) from scratch for end-to-end omni-channel journey optimization encompassing longitudinal-journey based pivotal moment discovery, multi-touchpoint attribution, behavioral pathway clustering, budget-constrained omni-channel intervention optimization, what-if-scenario based crystal-ball views - leveraging genetic algorithms, deep learning, causal-graph inference, and Hawkes point processes - in a big data setting for multiple verticals e.g. Technology, E-commerce, Investment Mgmt., Agriculture
• 2) Personalize.AI (SaaS Product) - Developed the end-to-end DS for a micro-segmentation based omni-channel item/offer recommendation SaaS product, leveraging advanced optimization algorithms (Dantzig-Wolfe decomposition methods) for large-scale optimization, constraint programming, Fourier transforms, time series clustering and causal meta-learners - for critical first-time deployments
• 3) AI For Insurance: Developed a novel Platform for Acquisitions and Attrition Management for a global Insurance giant, in challenging data conditions, leveraging deep learning and ensemble methods
• 4) AI For Investment Management: Developed ML Solutioning for a Marketing/Sales Optimization SaaS product - AI Guided Selling (AIGS)
• 5) AI For Airline Optimization: Developed Solutions for Business Development across varied airline prediction and optimization problems including planning, scheduling, disruption management and recovery - with a specific focus on turnaround time optimization; for one of the largest global airline companies

Feb 2020 – Apr 2020 · 2 mos

• Improved the external auction win-rate ML model for the bidding engine which drives Multi-Billion daily ad impressions for the programmatic AdTech ecosystem in the Healthcare domain processing billions of ad impressions daily

Feb 2018 – Sep 2018 · 7 mos · New York City Metropolitan Area

• Led and Oversaw the Full Spectrum of data science projects for one of the largest search engine advertising marketplaces with ~500 Million Daily Search Queries at TBs/hour data volumes, covering but not limited to:
• 1) Developed new Real-Time (<100 ms latency) CPC Bidding Algorithms for a disruptive search-engineering advertising product - Conducive Paid Suggest, leveraging mathematical optimization.
• 2) Developed Feature and Model Enhancements for Cost-per-click (CPC) prediction for both advertisers and publishers as well as Click-through rate (CTR) prediction
• 3) Developed a Multi-Armed Bandit (Thompson sampling, Upper Confidence Bound, etc.) based Marketing Budget Allocation Framework

Jan 2017 – Jul 2022 · 5 yrs 6 mos

• Providing mentorship to professionals and students at all levels, focusing on areas such as:
• Machine Learning (ML) and Statistical Modeling
• Algorithms and Software Programming
• Financial and Marketing Analytics
• Operations Research and Optimization
• Empowering mentees to develop advanced technical skills, enhance problem-solving capabilities, and apply data-driven strategies in diverse fields. Offering guidance tailored to individual goals and industry challenges.

Aug 2015 – Jan 2018 · 2 yrs 5 mos · New York City Metropolitan Area

• Director - Customer Knowledge and Strategic Insights (CKSI) for a company with ~3 Trillion USD of Assets Under Management (AUM)
• Drove Novel Advanced Data Science and Machine Learning initiatives for Fidelity’s Multi-Trillion Dollar portfolio, driving strategic omni-channel customer lifecycle optimization and profit maximization.
• 1) Reinforcement-Learning (RL) Based Adaptive Omni-Channel Recommendation Framework: Envisioned and Spearheaded the development of a self-learning RL framework, leveraging deep learning autoencoders, Q-learning with experience replay and target networks, to optimize omni-channel strategies to maximize long-term profits
• 2) Developed ML Models for Online Prospects to help drive Real-Time Online Marketing Personalization Strategies
• 3) Developed ML models (leveraging thousands of features) on Big Data for complex Markov-Chain transition matrix prediction to help optimize interventions for profitable customer segments (e.g. actively managed portfolios)
• 4) Built an interactive ML driven GUI tool for model agnostic what-if scenario-based driver contribution insights to help drive companywide model interpretability across diverse model classes
• 5) Developed an Empirical ML Roadmap for Business Optimization leveraging the state-of-the-art ML research in Spectral Clustering, Deep learning and Reinforcement Learning - leveraging both structured as well as unstructured data
• 6) Developed solutions for Operational Capacity Optimization for Non-Digital Channels

Feb 2014 – Aug 2015 · 1 yr 6 mos · New York City Metropolitan Area

• Led data science and analytics projects for Amex’s Multi-Billion Dollar Credit Card Acquisitions portfolios, driving Multi-Million USD impacts.
• 1) Global (US, South America, Europe, etc.) Modeling Lead - Global Corporate Payments (GCP) Acquisitions: Owned the suite of statistical models to drive acquisitions for the global GCP portfolio across many disparate countries and continents.
• 2) Outbound Telemarketing (OBTM) Modeling Lead - Small-Business (OPEN): Developed ML models to drive acquisitions leveraging external commercial data, external credit bureau data, in-house omni-channel temporal data - using Hadoop based distributed map-reduce implementations of ML algorithms
• 3) Merchant Risk Underwriting Strategies: Developed risk-management strategies by analyzing over 500 million records, for Next-Day Speed-of-Pay Initiative. Improved strategies to credit/fraud screen newly setup merchants. Analyzed strategies to diagnose fraud rings. Benchmarked discount rates for merchant negotiations using competitor data from third party companies.
• 4) Algorithmic ML improvements for Business: Achieved ~10% improvement using missing value imputation using Expectation-Maximization, Markov Chain Monte Carlo

Jul 2012 – Feb 2014 · 1 yr 7 mos · New York City Metropolitan Area

• Drove a Multi-Million USD reduction in net credit loss (of ~2 billion USD) by developing rigorous methodologies and advanced analytics to drive strategies to maximize contacts, promises and payments.
• 1) Risk Profiling/Score-carding/Rating: Leveraged risk scoring and segmentation models to create nuanced segments (e.g. straight rollers), by predicting roll-forward propensity between different delinquency stages
• 2) Credit Risk Modeling: Researched statistical techniques such as survival analysis, regression to predict accurately business-critical events (e.g. inter-stage transitions) for delinquent credit-card customers across various customer segments
• 3) Modeling for Omni-Channel Orchestration
• Built novel multivariate regression models to evaluate the contribution of every channel at every lag and optimize channel schedule to maximize EBITDA.
• Call-Intensity Optimization: Determining optimal daily calling intensities for different risk segments
• A/B Testing: Performed statistically rigorous analysis for A/B testing for channel strategies to evaluate business impact.
• 4) Portfolio Management Strategies for delinquent credit-card customers
• Leveraged CHAID decision trees to formulate rules for identifying profitable segments for dedicated assignment of accounts to agents for late-stage delinquent customers
• Agency-Commission: Developed linear-programing formulations to determine the optimal commission rates for external collections agencies
• 5) Operational Capacity Optimization:
• Accounts-to-collector ratio (ACR) optimization: Determining the optimal number of full-time collectors so as to minimize the NCL, considering metrics such as contact rate, saturation, and conversion rate.
• Performed Average Handle-Time benchmarking for operational efficiency improvements
• 6) Dashboard Presentations (to MD team) detailing end-to-end big picture portfolio-health metrics and MIS Design to identify key metrics for periodic evaluation and driving improvements.

Jul 2011 – Oct 2011 · 3 mos

• Holding office hours, Providing Technical Guidance to MBA students, Grading Assignments and Exams

May 2011 – Apr 2012 · 11 mos

• Result: 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐏𝐚𝐩𝐞𝐫 𝐢𝐧 𝐭𝐡𝐞 𝐌𝐚𝐭𝐡𝐞𝐦𝐚𝐭𝐢𝐜𝐚𝐥 𝐏𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐢𝐧𝐠 𝐉𝐨𝐮𝐫𝐧𝐚𝐥, 𝟐𝟎𝟏𝟓
• Contributed to Improvements on the State-of-the-art Results on Multi-Row Cutting-Plane Methods in Integer Programming:
• The corner polyhedron is described by minimal valid inequalities from maximal lattice-free convex sets. For the Relaxed Corner Polyhedron (RCP) with two free integer variables and any number of non-negative continuous variables, it is known that such facet-defining inequalities arise from maximal lattice-free splits, triangles and quadrilaterals. We improve on the tightest known upper bound for the approximation of the RCP, purely by minimal valid inequalities from maximal lattice-free quadrilaterals. We generalize the tightest known lower bound for the approximation of the RCP, purely by minimal valid inequalities from maximal lattice-free triangles, to an infinite subclass of quadrilaterals.

Jan 2011 – May 2011 · 4 mos

• Surveyed Decomposition Techniques and Constraint Propagation to Solve the Two-Dimensional Bin-Packing Problem

Jul 2010 – Jun 2012 · 1 yr 11 mos

• Studied and Researched the State Of The Art in Linear/Nonlinear/Integer/Constraint Programming, Graph Theory, Networks and Matchings and Discrete Optimization

Jul 2008 – Apr 2010 · 1 yr 9 mos

• 1) Led a team to generate business insights for a Prescription Spillover Phenomenon in the Managed-Care US Healthcare ecosystem using rigorous statistical regression analysis (handling aspects such as Mahalanobis outlier removal functional-form transformations, error auto-correlation, homoskedasticity, multi-collinearity, etc.) and time series ARIMA modeling - for 10 drugs of a leading pharmaceutical company in the US market. Quantified the doctor prescribing behavior using trends showing the sensitivity of market share of the client drug to share of practice of the doctor in different categories of plans based on step-therapy, copay etc.
• 2) Optimized Promotions and Discovered Marketing Channel Attribution leveraging in-house tools based on advanced Econometric Modeling
• 3) Sales-force optimization: Determined the optimal size of the sales-force teams for leading US pharmaceutical companies, by determining optimal promotion territories using metrics such as workload balancing, geography, connectivity etc.
• 4) Co-promotion Effectiveness Analysis: Analyzed the co-promotion effectiveness in the US market for the same drug by two different pharmaceutical companies, to determine the optimal promotional strategy
• 5) Marketing Effectiveness Analysis: Performed Statistically Rigorous test/control analysis for marketing promotions
• 6) ROI analysis: Determined the ROI on different promotion channels (on-site samples versus. phone calls versus medical conferences versus on-line marketing etc.) for leading U.S. pharmaceutical companies

Jun 2007 – Jul 2008 · 1 yr 1 mo

• Result: Published sole-author research papers in:
• 1) 𝐈𝐄𝐄𝐄 𝐈𝐧𝐭. 𝐒𝐲𝐦𝐩. 𝐨𝐧 𝐀𝐝𝐚𝐩𝐭𝐢𝐯𝐞 𝐃𝐲𝐧𝐚𝐦𝐢𝐜 𝐏𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐢𝐧𝐠 𝐚𝐧𝐝 𝐑𝐞𝐢𝐧𝐟𝐨𝐫𝐜𝐞𝐦𝐞𝐧𝐭 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠, 𝐔𝐒𝐀, 𝟐𝟎𝟎𝟗 and
• 2) 2009 WRI Global Congress on Intelligent Systems, China
• Objective: Developed novel actor-critic algorithms for reinforcement learning (RL) in two-timescale actor-critic algorithms using long-run average-reward criteria and linear value-function approximation - to minimize the high variance in policy-gradient estimators.
• Key Contributions:
• Introduced a stochastic-gradient-based critic update to improve policy-gradient ascent.
• Developed a novel baseline structure for variance minimization
• Derived a novel actor update leveraging this optimal baseline for an existing RL algorithm, and further introduced an unbiased policy-gradient estimator based on the policy-gradient theorem with function approximation.
• Provided a variance-minimization-based interpretation for existing algorithms

Jan 2007 – Jan 2008 · 1 yr

• Setting and grading assignments
• Providing Technical Guidance to MTech students
• Courses: Markov decision processes, Deterministic models of Optimization

Jan 2007 – May 2007 · 4 mos

• Intelligent agents in Supply Chain Management:
• Researched state-of-the-art in Reinforcement Learning, bidding, and mediated constraint relaxation in optimal scheduling, optimal ordering decisions in multi-echelon inventory systems.

Jan 2007 – May 2007 · 4 mos

• Tech-Fest (Asia's largest science & technology festival) Management System:
• Designed entity-relationship diagrams, database schema, multilevel data flow diagrams, activity diagrams, use-case scenarios and deployment diagrams.

Aug 2006 – Dec 2006 · 4 mos

• Optimized Signal Times at a Major Road Intersection using Simulation-based Optimization:
• Collected real traffic data at different times of day/night. Performed detailed traffic modeling and simulation-based optimization in Arena and proposed change in green-signal times.

Jan 2005 – Jan 2006 · 1 yr

• Project member in Center for Indian Language Technology Solutions (CFILT) - Dept. of Computer Science and Engineering: under Professor Pushpak Bhattacharyya
• Result: Published a 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐏𝐚𝐩𝐞𝐫 𝐢𝐧 𝐭𝐡𝐞 𝐉𝐨𝐮𝐫𝐧𝐚𝐥 𝐈𝐄𝐓𝐄 𝐓𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐑𝐞𝐯𝐢𝐞𝐰, 𝟐𝟎𝟎𝟔
• Machine Translation: Agro-Explorer - a multilingual, meaning based search engine in the agricultural domain first extracts the meaning of the query and then performs a search based on this extracted meaning. Hence, search can be carried out even if the language of the query is different from the language of the documents. The meaning is represented in the form of semantic-graph based Universal Networking Language (UNL) Expressions. The search is carried out using UNL expression matching. The relevant documents are in the UNL form. The Deconverter converts these documents into the language of the user's choice. We developed the Deconverter, with Marathi as the target language, for Agro-Explorer. The deconversion proceeds through the following four stages: a) Syntax Planning (Semantic UNL Graph traversal) b) Lexical Replacement; c) Case Mark Insertion; d) Morph Generation.