Moinuddeen SMRK

• Ensuring customer delight with improved product offering & best in class features. Currently developing a machine learning model to determine propensity of user reactivation using historical user on-app behavior & other characteristics to help improve daily reactivation, better targeted campaigns & optimizing campaign driven expense.
• Have put Leading efforts in cross-functional collaboration and stakeholder communication and responsible for product development from inception to implementation
• Worked with tech team, compliance and design to develop user reactivation campaigns and proposed inclusion-exclusion criteria for optimal cohort selection for campaigns leading to a 25% increase in monthly reactivated users
• Proposed a product enhancement to encourage user transaction (Add money) in the wallet leading to 8% increase Add-money transactions on monthly basis.

Pricing

• Improving orgs financial gains through effective pricing leading to customer and merchant happiness Owned the execution of training, testing & deploying new configuration of pricing models as switchback experiments for new service areas for the transport wing (4W) with continuous monitoring of model performance in terms of keys metrics such as gross take, conversion metrics in order funnel like EBR, BCR ECR etc
• Built Looker (Data Studio) dashboard to monitor pricing metrics and verse with google cloud platform (GCP) and Vertex.ai Have been the PoC for evaluating BigEye, a data validation platform and created alerts for all tables managed by the team and owned several data pipelines and resolved any issues in a timely manner
• Built a Looker (data studio) dashboard for monitoring daily, weekly and monthly trends of key metrics related to pricing
• Additionally maintained several scheduled pipeline in Airflow
• Monitored & Maintained multiple A/B tests at a weekly basis
• Performed a statistical analysis to ascertain network effect observed in a service area to understand Driver & User behaviour in terms of positioning, pricing and trip distance distribution causing under supply in certain areas
• Performed several quick analyses to help out business teams to understand past and expected future behaviour of the market.

Serviceability

• Improving order life cycle & customer experience using predictive modelling to estimate serviceability based on driver availability Co-developed a serviceability model that predicts driver availability (or Driver not found) for orders using a XGboost model increasing Monthly savings by
• In collaboration with stakeholders, understood the impact of risks associated with false positives and false negatives in the model and developed a custom loss function to tune the Xgboost model to asymmetrically penalize false positives and false negatives in outputting serviceability of an order
• This helped the model better understand the risks of false positive (losing merchant commission) and false negative (merchant compensation) which lead to 2% increase in monthly savings
• To combat imbalanced training data, used threshold calibration for the model to increase the chances of probabilities to match the true likelihood of events
• Further improving the serviceability model, used Facebook’s prophet model to forecast near future serviceability to ensure customer delight through proactive, soft warnings
• Built a Looker (Data Studio) dashboard to monitor the daily & weekly model performance and financial metrics

Natural Language Processing

• Redefined unstructured data analysis from clinical free text through natural language processing, semantic & time expressions using ontologies from Humedica, OPTUM & UMLS to help reduce 90% time consumption
• Owned the execution from conceptualization to post implementation and led a team of 4 to build a data solution (Plug and play) to reduce ~90% time consumption in data analysis & review process using un-structured data
• The tool was presented in the client org's Annual R&D symposium and garnered recognition from several teams Developed a re-usable automated framework for extraction of clinical named entities augmented with a time expression analyzer, semantic association & negation handling using Python, & cTAKES
• The solution was endorsed by the Safety Science Review Board, an independent committee for global org- wide use of innovative solutions

R&D Advanced Modeling

• In collaboration with biopharma R&D SMEs, re-imagined drug manufacturing problems of non-linear scale-up through digital simulation of Bioreactor equipment Developed a predictive model with over 85% accuracy using a combination of equation based modeling, agent based modeling & computational fluid dynamics for digitally stimulating bioreactor environment to test previously untested conditions and frontload causes of failure
• Optimized with parallel processing algorithms & cloud infrastructure Presented the unique approach for Bioreactor Simulator in the 2nd National Conference on Multidisciplinary Design, Analysis and Optimization ‘19

Retail Analytics

• Measured the effectiveness of a promotional campaign in physical stores by performing lift-analysis in consideration with the halo effect, cannibalization & pull-forward effect
• The lift was also compared with control & test stores identified using K-nearest neighbours classification