Kuona is a Software as a Service platform utilizing deep learning techniques to optimize product pricing, promotions and inventory management for multinational companies in the USA, Latin America and Europe. During my tenure here, I contributed to the following projects:

Promotions. Solutions for analyzing promotion outcomes (post-mortem) and predicting future promotional performance.

• Participated in migrating ETL processes from PostgreSQL and Django commands on Lambda to a Data Warehouse architecture using Glue Catalog, Spark jobs and a Medallion architecture with S3 buckets (bronze, silver, gold), orchestrated via Airflow. This migration improved execution time by up to 70% and reduced costs by approximately 30%.
• Built exploratory data analysis dashboards using QuickSight and Grafana, querying data directly from our S3 buckets to identify sales increases and cost-of-goods reductions. This initiative eliminated the need for bi-weekly meetings dedicated specifically to this analysis.
• Collaborated in initiating and designing an internal automated testing project to enforce client-specific data contracts, data type requirements, and business rules during data ingestion. The system generated automated violation reports sent directly to stakeholders.
• Implemented algorithms to calculate promotion KPIs and trained neural networks using cloud services such as Lambda and EC2 instances.
• Responsible for creating, monitoring and reporting on data pipelines and insights across various client projects.

Pricing. Solution to evaluate price elasticity based on metrics including cost of goods, competitor pricing and regional factors.

• Played a key role in onboarding the first client onto this solution, overseeing data integration, designing data models and implementing monitoring and business rule enforcement.
• Designed, orchestrated and maintained ETL pipelines for ingesting data and running customized deep learning algorithms.
• Delivered in-person client presentations explaining the solution, its functionalities and the data requirements necessary for optimal workflow.

Perfect Order. Solution for inventory flow for specific location according to the predicted demand.

• Collaborated closely with Machine Learning Engineers to create tailored feature engineering solutions per client and automated their integration within the company’s deep learning framework.

Tech stack used: SQL, Python, Django, Kafka, Spark, Airflow, AWS (Lambda, S3, Glue, Athena, EC2), PyTorch, PostgreSQL, Redshift.