Date: Jan 2022 - Feb 2023
Roles: Data Scientist, Data Engineer
This project was completed while I was a consultant at Melio AI, and some of the project details have been obfuscated.
Overview
Worked on various machine learning forecasting projects, data engineering ETL pipelines, and visualization dashboards, along with general software engineering in Python.
My Responsibilities
- Equities and FX price forecasting:
- Reviewed existing implementation of ML models.
- Investigated forecasting accuracy of statistical and ML models (ARIMA, MLP, LSTM, RNN) using sentiment and historical price data.
- Benchmarked ML model forecasts to simple models.
- Built Flask API to allow traders to query models and integrate forecasts into their workflows.
- FX price formation:
- Investigated USD/ZAR price formation to balance inventory depletion and profit maximization.
- Implemented features of price formation models into production using containers and streaming data.
- Designed and deployed a visualisation dashboard in PowerBI displaying FX rates and metrics.
- Client behavior modeling for CRM:
- Investigated improved methods for modeling client behavior around canceled/unfulfilled FX trades.
- Built an ETL pipeline to process and transform internal data sources into client behavior metrics.
- Created a custom RFM model and client tier ranking system.
- Stakeholder communication and knowledge sharing:
- Communicated designs and experiment results to technical and non-technical stakeholders.
- Documented new features and simplified statistical concepts for better understanding.
- Conducted knowledge-sharing sessions on machine learning and deep learning fundamentals.
Outcomes/Impact
- Reduced costs by eliminating infeasible/inaccurate forecasting models.
- Enabled traders to visualize and utilize FX pricing model outputs effectively.
- Identified areas of intervention with a custom client ranking system to reduce trade cancellations.
Tools Used
- Python
- TensorFlow
- Scikit Learn
- Flask
- Azure
- PowerBI
- Docker