Nac Stoklas

I helped build a fully distributed Azure cloud Databricks-based solution for retailers.

The solution segments items into different categories and makes predictions for scheduling optimal times considering deliveries, current stock levels, vendor preferences, and cargo requirements like shipping and truck space.

I built most of the scalable data pipeline and made prediction models. I also developed most of the supporting DevOps pipelines for CI/CD.

I've used cloud technologies such as Azure functions, Data Lake, Blob storage, Databricks, and others.

I built an API developer portal using Azure API management services.

Businesses use it daily to get high accuracy, low latency forecasts for their business needs. I also built several APIs on offer, including multiple methods for different types of forecasts, such as weather-based, window methods, statistical methods, and boosting-backed methods.

I built a basket analysis case using association rules and an exploration dashboard based on the basket analysis to optimize item placement within stores, offer better deals and find items that sell well together.

I also built a custom extension that allowed for write-back functionality within Qlik Sense, used for the dashboard.

I built a connector that transforms a common graph database, specifically Neo4j concepts and hierarchies, to the relation database format used by all major BI tools.

It was integrated with Qlik Sense for a major Slovenian retailer to analyze their graph database data using standard BI tools.

I built an Azure CI/CD pipeline that supports project development for a whole ML team for one of their core products.

It automatically tests the code, builds artifacts, and deploys them on production clusters (Spark and Databricks), updating libraries and code when there is the least customer traffic.

I built the ETL for evaluating the business performance of integrated systems and the dashboards required for generating business insight that helped improve related systems, including stock management, warehouse management, and cargo.

I've ported the QlikView application with a complex ETL and dashboarding to Qlik Sense.

It entailed finding equivalents for many major QlikView features in Qlik Sense, adjusting the ETL so that it finds into the Qlik Sense script editor, and adjusting the visuals to fit into the newer engine.

I've built an API that can be used for multiple and diverse types of time series forecasts.

This supports many parameters, such as weather and holidays (for ISO country names) and more fine-tuning for automated parameter selection. It can be easily embedded in major ERP solutions—FO, BC, NAV, and Infor. It offers support for statistical algorithms, boosting methods, and classical ML.

I've helped develop a cheap camera's CV object detection of a ski jumper. On top of this, I've worked on automatically calculating the jump distance using OpenCV, and I've worked to port this functionality onto a live video stream.