Data Scientist2019 - PRESENTAltoros
Technologies: GIS, Kubernetes, Docker, Flask, FastAPI, Neo4j, Keras, TensorFlow, Python
- Constructed a proof of concept (PoC) for field crop detection and segmentation, tracking crop health status, and detection of natural hazards.
- Built a PoC for automating PDF processing functions with machine-learning algorithms that detected structural elements, extracted text, tables, graphs, and table of contents (ToC) from documents with a complex layout.
- Developed microservices for drivers' behavior analysis, route map matching, and trip summary calculation.
- Created a model for route and destination prediction based on the Hidden Markov Model.
Data Scientist2017 - 2018Kreo Software
- Developed a set of tools for Amazon Mechanical Turk (MTurk): a framework for publishing tasks, validating work, and selecting and training workers.
- Supervised several MTurk projects with challenging NLP tasks that required a rigorous screening of workers.
- Improved the classifier of construction elements in the Uniclass system.
- Optimized the classifier for production and deployed the trained model.
Data Scientist2016 - 2017Kinross Research
- Developed an entity extraction model for extracting dimensions, squares, volumes, ranges, and so on with corresponding units from a textual description of construction elements. I built two versions of the model: a rule-based version and a version based on a recurrent neural network.
- Created a data-specific PDF table extractor and integrated the extracted data into the knowledge base.
- Built a web application: a type of annotation tool for labeling data for ML models and for the core project engine.