Hesham Mahrous

RTA - Dubai: Built a sentiment analysis tool for CRM and social media data to show insights and show analytics on road conditions and services all over Dubai. The tool helped me management to gain insights and very quickly take actionable insights The potential outcome is to increase customer satisfaction and increase process efficacy. In addition to that worked on parking predictions occupancy using state of the art time series machine learning techniques which lead to better urban planning.

● Using IOT data parking sensors created a POC and implemented production code deployed using docker for a company called Open Spot in Vancouver Canada. The prediction model is an API that takes a get request with a json payload of user demographics information and historical events and outputs a prediction of most likely area of parking of the user and a best optimized recommendation scenario for the user to follow.
● Using card loyalty data, sales and products attributes data built a segmentation machine learning algorithm that categorizes customers and gain better understanding of their purchases and shopping behavior for a retail store.
● Using LSTM, built a forecasting algorithms that predicts the weekly sales of each item in the store with very high accuracy, the model is adaptive to seasonality. Using a similar technique forecasted the user’s next visit to the store with very high accuracy as well.
● Using marketing and promotional data, products attributes, weather and macro-economics data such as CPI and PPI index, built a machine learning algorithm that created a “what if” scenario to predict the outcome of marketing and advertising campaigns on products.
● Helped the technical team of a health insurance company built a fraud detection system to detect and flag fraudulent claims, the outcome of the system saved millions of dollars on monthly basis as it turned out the system correctly flagged 5% of the claims.

Conducted Customer Life Cycle Analysis, Churn rate analysis, predictive modeling for marketing campaigns and promotion optimization and predictions through big data. By finding some behavioral patterns from the user, communicated the insights and suggested ideas that helped increase our google play ratings from 3.1 to 4.6 stars in weeks. Some other insights has led to the increase of transactions by 40%.
● Using LSTMs applied Natural Language Processing problems, such as sentimental analysis and document classification analyzed online reviews of different languages. Extracted meaningful keywords, topics, association analysis. The results of the analysis helped product managers and cooperate take business decisions for operations, and R&D projects by giving them actionable insights from the data.

● Using Java and Java Script developed a novel text mining platform used for nomenclature searching to match for imprecise data, achieved accuracy scores higher than 95 %.
● Accepted for US patent, first author and main contributor of this invention. Currently targeting Journal conferences for publication.
● Using data optimization and machine learning developed a Cost Analysis formulation to inference cost patterns of open source software used in Business Intelligence products. Currently applying for a second patent and then conference publications. The outcome impacted the management to restructure the resources of the organization.

Researched Automated Speech Recognition (ASR) noise environment problems, then suggested machine learning algorithm to implement. It increases speech recognition accuracy scores.
•Using Matlab I researched and developed a machine learning algorithm to de-noise speech in different noisy environments which improved the Automatic Speech Recognition accuracy by 40%.

Developed a machine learning algorithm based on Gaussian Mixture Models for vein localization based on fuzzy infra-red data, this work achieved an exclusive technology license and I was honored an appreciation award.
•Maintained the vein locator handheld device using C, Cypress microcontroller PCB. Achieved high accuracy localization scores in which it passed the required clinical trials for health Canada and FDA approval.

Developed and tested a firmware code for a PIC micro-controller and a PCB. The product was developed to monitor and control gas sensors that use SPI and Modbus communication protocol. Achieved and developed this product in one month.