🤖Project: address the problem of attention drop-off in online ads. The product has two main components: the ad pushing at scale component and the creative component. Our clients create
ads using dynamic components that can have different values when
being pushed to social platforms.
📈 Contributions:
- attention drop-off algorithm
- creative feedback on attention driving elements
- increased client engagement
- better exposure from the AI community leading to better exposure for the company.

🤖Problem: when collecting and processing metrics data from Facebook for the ad pushing platform there were frequent changes in the metrics returned by Facebook, which caused spending between 2-4 development hours each time we updated our SQL database to accommodate new columns. This led to delays in updating clients with the latest metrics data, resulting in the loss
of valuable insights and opportunities for business growth
📈 Contributions:
- proposed and implemented a Neo4j graph database that streamlined the process of collecting and processing metrics data. This solution not only saved significant development time but also enabled us to collect new metrics data more efficiently, providing our clients with the latest insights they needed
to make informed decisions
- proposed and developed a persona discovery tool that utilized Facebook metrics data and non-negative matrix factorization to identify the different types of users viewing our clients’
videos. This tool provided our clients with valuable insights into
their target audience without accessing individual personal data,
enabling them to create more granular ad personalization and drive
better business results
- result of these solutions, our clients were able to access and
leverage the latest metrics data more efficiently, enabling them to
make informed decisions and drive better business results. This
contributed to the overall growth and success of the company, enhancing the company's reputation as a leading provider of innovative ad pushing solutions

🤖Problem: the platform - is a client-facing application. The client-service team would aid the clients in using it and pushing ads on different platforms (e.g. twitter, facebook etc.). However, we discovered a significant bottleneck: wrong formatted input (e.g. errors in text, videos too short) would be discovered only after a push, which would then fail to validate on the social platform side, delaying the
go-live of the ad. This, in the ad industry, is a critical problem - as certain ads need to go online at certain hours (e.g. during a foot-ball game ).
📈 Contributions: , I implemented the first automatic ads-pushing pipeline, which included early validation, becoming the standard template for future clients. This pipeline improved the
efficiency of the client-service team and brought more clients, with time-sensitive ads