Lead a cross-functional (Data Scientists, Engineers, Annotators) team in Data Science, Data Platform, Personal Finance Management, and Risk Tech initiatives at Jupiter.Data Science initiatives are across the areas of Personal Finance Management, Customer Service (like chatbot), Risk Tech and User Profiling.Personal Finance Management, empowers users with tools and insights to better manage their finances, including spend categorisation and tracking, recommending actions for improving financial health.Data Platform, implementing a data lake house strategy to improve data accessibility, reliability, performance and reduce total cost of ownership, along with key focus on data access controls and governance.Lead the Risk Tech initiatives, specialising in credit risk modeling across various products and transaction fraud monitoring. Implement credit risk models to assess the creditworthiness of customers, optimizing loan approval processes and minimizing default risks. Transaction fraud monitoring systems, leverage real-time data analysis and ML models to identify and prevent fraudulent activities.

Established and led the Data Science team- DS team had a footprint with multiple deployments and impact in the areas of Fraud and Abuse, Supply Chain, Search, Recommendation, Ads, Localisation, Forecasting and Planning, Pricing, Catalog, Customer Service, Sales force optimisation- Deeply involved in all the DS team projects from problem formulation to data requirements to modelling strategies to evaluation and deployment along with some hands-on ownership of couple of projectsEstablished and led the Data Platform (DP) tech team- Conceptualised and built in-house DP with a data lakehouse strategy- Widespread adoption of DP with ~1 Pb data footprint, 20000 daily queries, 2500 dashboards, 800 data sources ingested, 150 DS/ML daily jobs across reporting/dashboarding, analytics and data science use cases- Reduced total cost of ownership of DP by roughly 75% over 3 quarters without degrading user experience via multiple initiatives leading to significant backend improvements and usage optimisation- Established ML Platform sub-team within the DP team with successful closure in MLOps charter with in-house Feature Store, ML model tracking, standardised ML model pipelining, deployment and monitoring. Multiple projects live via the MLOps initiatives including real-time ML deploymentsEstablished and led the Product Analyst team- Set up the product analyst charter along with role clarity and expectations. Hybrid team structure with central reporting but distributed working within the respective product tech team- Several initiatives driven centrally to standardize AB experimentation framework, core data views, L0/L1 metrics definition, company-wide training of larger analyst communityEstablished and led the Fraud and Abuse team (DS, Engineering & Tech-ops)- Successfully setup DS led Fraud and Abuse charter in solving Fraud and Abuse problems across all stages of e-commerce from identity resolution to fake/risky orders to RTO/return prevention

Was the first hire in the Data Science leadership team at Flipkart, thus was privileged to help build the DS team, competencies and processes. Worked across multiple domains of Search, EKart, Pricing/Planning, CX and Fraud detection. Led the Search and EKart DS charter. Deployed multiple modules across Search around query understanding, ranking, personalisation and autosuggest. Seeded and defined the EKart DS charter.

Big Data@Health- Trying to tease apart the hype of Big Data from their true applicability to Healthcare domain (Providers,Payers,Pharma & Patients-4P's of healthcare)Healthcare Fraud, Waste, Abuse & Error:- Understanding the landscape of Fraud, Waste and Abuse in commercial and federal(medicare)/state(medicaid) insurance as well as applicability of analytics techniques in detecting and preventing Fraud, Waste & Abuse.

- Claims Rework Analytics tool: Claim payment errors and rework drive a significant portion of the administrative costs and service issues of health plans. Accenture's Health practice searched for new technology that could to reduce claims rework for our clients - not finding a solution, they requested Accenture Technology Labs to help create one. We created a claims rework identification platform using predictive modeling (machine learning) techniques. Our testing over multiple client pilots indicated that the engine can identify the majority of future rework claims and score them effectively enough to allow for practical prepayment intervention.- Hands-on lead for research and development of the Claims Rework Analytics asset and led it through the proof of concept and pilot engagements of the asset with the client teams. I have been involved in all aspects of the tool from research conceptualization, developing the system, interacting with client teams, analyzing the results and translating it into business case. I have also helped in successfully transitioning the asset from Labs to industrialization team for client deployment and continue to support the asset as a subject matter expert and continued R&D.Proposal Assistant tool: - An assessment of the proposal writing processes as executed within Accenture, including the challenges and opportunities for improvement. Identify ways we might address high-priority issues through the application of knowledge management technologies in innovative ways.

- Worked on the Briefing Assistant specialist under the RADAR project (http://www.radar.cs.cmu.edu/). In particular, I was interested in the problem of summarization which involves abstracting information from the given text, for e.g. a set of emails, or a set of events in the form of a concise briefing. We developed a report writing system that summarizes task-based event information and uses templates for presenting the report to the user.- It was a great learning experience as I was the point of contact for our group. I led our research, development, integration and testing initiative which involved extensively interacting with other groups in RADAR.

- Worked on building an Indian English Text-to-Speech system. - Explored the implication of speech technologies on pervasive devices, addressing the problem of speech application adaptation by dialog call-flow reorganization for pervasive devices with different memory constraints. - Worked on building a continuous speech, speaker-independent Hindi Speech Recognition (Desktop application) system. Main contribution had been to develop a novel technique of generating phonetic spellings of words to build the vocabulary in the system. Also designed and built the feature of adding words to the vocabulary software.