Jul 2024 – Jan 2025 · 6 mos · Bengaluru, Karnataka, India · Remote

• Developed & supported payee risk models for victim in APP fraud, saving £80K £ by detecting and preventing fraudulent activities.
• Designed and deployed an end-to-end ML system for risk modeling, integrating Fiddler AI for drift detection and model monitoring.

Jun 2022 – Jul 2024 · 2 yrs 1 mo · Bengaluru, Karnataka, India

• As a Lead data scientist at Paytm's digital advertising & Travel vertical that offers innovative solutions .
• Paytm- Travel
• Built and led a comprehensive data science team in Paytm Travel, set up data science infrastructure, and built pipelines for feature engineering, training, deployment, and A/B testing frameworks. Designed the ML system for handling 60 million customers.
• ML solution to search rank all the relevant buses for the user based on their behaviour. Trained ML algorithm & provided ranking predictions for 25 million customers, encompassing 0.35 million buses. .Performed AB testing & Increased booking conversion by 20%.
• Cancellation Protect Flight: A ML Solution that predicts if a user is eligible for cancellation protect insurance while booking the flight & once after the eligibility also predicts the price of the premium amount. Impact : Turned a loss-making vertical to a profitable one (1 million$ per month)
• Paytm Ads
• Designed a model for audience targeting that segments and targets users similar to those who are high-performing within a given advertiser or category, with the aim of improving campaign results and advertiser return on investment (ROI); ML segments demonstrated approximately 60% higher CTR compared to rule-based segments and increased the platform's CTR by 15%.
• Identified customer personas and created cohorts based on behaviours observed on the Paytm app and ads; this initiative provided insights into hidden patterns and differentiated between good/average and active/dormant customers both for first-party data and MOA.
• Developed a generative AI (GenAI) proof of concept (POC) for revenue uplift by enhancing an ML model with visual creative and advertiser features using GPT-4 text and vision API; the POC saw an up to a 13% performance gain, which led to significant impressions inventory savings and a potential ~5% monthly revenue increase.

Mar 2020 – Jun 2022 · 2 yrs 3 mos · Bengaluru, Karnataka

• Operated as a Data scientist at a company that is recognised as the global leader in Industry 4.0
• Historical Time to Action: Developed a model to predict the time to take predictive maintenance action for machine related issues using inference modelling. Significantly reduce the resource cost saving by 50%.
• Batch Analytics: Developed a product for managing & predicting batch production in the FMCG sector which will help in forecast production planning & resource allocation. Successfully deployed the ML system for world leading FMCG company.
• Proficy Csense : Developed an Industrial advanced analytics software for SME with No code Features and presented post sales demo for more than 20 global customers.
• Maintenance Interval Optimization : Used optimization algorithm to determine the proper predictive maintenance interval to avoid excessive or insufficient maintenance, which also help in reducing cost & risk.

Apr 2019 – Feb 2020 · 10 mos

• ⦿ Build an end-to-end analytical product for steel industry for open casting.
• ⦿ Developed a new algorithm to change the classification threshold of ML models using mixed-integer programming which help us to make models for highly imbalanced datasets.
• ⦿ Developed a failure prediction classification model for steel casting using machine learning models & sucessfullys deployed & setup MLOps system with drift detection features on real-time prediction
• ⦿ Developed a regression model to calculate the convection heat transfer from the steel & using those features build a dynamic threshold model.
• ⦿ Created a object detection model to capture the issues on the blade of the wind turbines.
• ⦿ Successfully deployed the model with scalable architecture & reduced corrective/reactive maintenance cost by 2 percent of the revenue.
• ⦿ Created CI/CD pipeline with automated performance testing and memory profiling for ML-based algorithms
• ⦿ Created Maintenance Interval optimization: Used optimization algorithm to determine the proper predictive maintenance interval in order to avoid excessive or insufficient maintenance,

Jun 2016 – Feb 2019 · 2 yrs 8 mos · Bengaluru, Karnataka

• ⦿ Build an end-to-end analytical product for the US govt to investigate child abuses in the country.
• ⦿ Created an analytical tool for in-depth analysis of location-wise crimes with the unsupervised algorithm.
• ⦿ Created a machine learning model which tried to predict the severity of the case with the past data.
• ⦿ Worked on POC for image captioning for the child portal module.
• ⦿ Created a review system for the Client portal using NLP & deployed the model on the server which helps the team to make the product in a better way for future sprint iterations.
• ⦿ I have worked as a member of the LineSight Data Analytics team of Unisys LineSight products. LineSight is a software for doing intelligence analysis on travelers and air cargo shipment, which reduced asset maintenance costs by 40 percent & mitigated risk.