Site reliability and perf.

Productionizing ranking models used in recomm scenarios across MSAI and FAST orgs which serve (or will serve) 100M+ MAU. Dedicated to constant improvement of MLOps standards and practices for FAST Norway org.

Worked on an intelligent scenario which leveraged eyes-off datasets available in Microsoft Graph. Data engineering, model training.

Designed and developed an offline inference pipeline used to backfill company’s used car data inventory through application of classification, object detection and OCR models (alone or in a combination). This pipeline is fully configurable, optimized for cost saving, automatically scales depending on the workload, and serves over million requests daily while simultaneously supporting 11 models. Supervised and managed a team of data quality analysts which continued to train and validate CNN models.

Software developer on several projects based on Spring Framework, utilizing some of the most popular big data technologies- from NoSQL storages like Cassandra, Redis and Elastic to data analysis and prediction solutions like Hive and Spark.

Solr Developer on a local search product based on Apache Solr and Apache Nutch; worked on crawl, index and search. Customized and deployed on 8 different sites. Software Engineer on a solution based on Liferay- an open source Enterprise Portal. Built on Spring Framework which utilizes Solr for product search.

Worked in the context of Monnet FP7 project on indexing and search. https://nuig.insight-centre.org/unlp/people/alumni/srdjan-prodanovic/

Worked on telecommunication administration system for Mobilkom Austria AG. On the front I have worked in Adobe Flex, and in the back Spring and Hibernate for DAO layer.

Worked on an e-advertising campaign management system for Italian Yellow Pages. Worked on integration of keyword expansion module in the system and on overall system's backend using Java EE on Myslq DB. Written custom spiders and crawlers for obtaining content of Yellow Pages.