• Neural Network model for identifying and flagging duplicate driver photos in order to prevent fraudulent activities
• Anomaly detection UI + alerts: a user interface that facilitates the detection of anomalies in traffic, features, rules, or models. Implemented an alert system to notify users in real-time when an anomaly is detected.
• Model’s interpretation: a user interface to enable Product and Business teams to interpret model outputs. The interface provides transparency and facilitates model validation.
• Pre-auth model: a machine learning model to minimize credit card underpayment losses. The model employs advanced techniques to improve accuracy and minimize false positives.
• Fund outflow model: a machine learning model to mitigate the risk of fraud associated with fund outflows from the Careem platform. The model incorporates sophisticated algorithms and statistical techniques to identify potential instances of fraud and minimize financial losses.

• Developed and implemented an AB testing framework from scratch for the Integrity team. This facilitates A/B testing for two sets of rules and enables comparison of their respective outputs. Subsequently, the Data Scientists/Data Analysts can make informed decisions about whether to deploy or remove these rules, and gradually turn on traffic for them.
• Automated measurement of precision using experimentation and test/control groups, which was developed from the ground up. This process enables a predetermined percentage of traffic to bypass Integrity controls, and measures the precision of these controls based on user status over time
• Automated various human processes, such as creating an auto clawback process to retrieve money from blocked customers and captains, unblocking inactive captains upon request from Operations team, and blacklisting cards to prevent fraudulent activities
• Created a card addition flow, which includes logic and rules for adding cards. Depending on the card feature, the flow allows for smooth addition, 0$ authorization, 3DS verification, or card decline.
• Designed and deployed a Chargeback model to reduce losses from Chargebacks. Based on various user features, the model determines whether to call 3DS during the TopUp process. The model was later extended for Mobile Recharges, resulting in a significant reduction of fraud and losses
• Implemented real-time payouts Integrity checks for Captains by creating controls for instant captain payouts.
• Developed a Captain Payout model to automate daily payouts for Captains.
• Managed the Destination Change project, which entailed developing and deploying the necessary features for users to modify their trip destinations.
• Designed and implemented a Trip Completion model that includes rules and a model to block captains and customers in real-time once a trip is completed.

• Developed content for Data Science course. Prepared lectures and materials for topics: Feature Engineering, Ensembles, Gradient Boosting, Time Series

Data Science lecturer, topics including Classification, Regression, EDA, Unsupervised learning, etc. Github projects

• Video recordings can be found by the link: https://bit.ly/3csuc0g
• Course materials can be found by the link: https://github.com/DmitriiDenisov/mlcourse_dubai
• Community page: https://www.meetup.com/Dubai-Data-Science-Meetup/

Participated in several PoC projects, including those related to neural networks and satellite images

Developed recognition models based on neural networks, and also developed a graph of companies inside Russia to detect corruption schemes during tenders

Lecturer of Introduction into Deep Learning

• Course materials can be found by the link: https://www.coursera.org/learn/intro-to-deep-learning