Claudio Hurtado Sepúlveda Email & Phone Number

Atlanta, Georgia, Usa

Assisted with the development of Computer vision models, working on object detection and on image segmentation. Worked in the experimentation phase with YOLO models, improving an initial score of 37% mAP up to our state of the art of 82% mAP, when detecting relevant construction objects in plans.Developed instance segmentation models with PadleSeg, achieving 90,2% IOU and developed its training pipeline using kubeflow in GCP,Assisted in the deployment of said models using Nvidia Triton as backend, developed training pipelines using Kubeflow, set up docker containers and generally experimented with different techniques, such as augmentations and post process to boost model performance, all while using GCP.In NLP, finetuned and prompt-tuned PaLM models from the GCP cloud and quickly developed NLP models thatcould replicate human data retrieval with 80% accuracy.Trained information extraction models using gemini-pro and prompt engineering, quickly achieving results with 92% accuracy.

Provincia De Buenos Aires, Argentina

✔️ Used NLP to analyze legal documents. Created a NLP (roBERTa) model to determine the duration of each judicial step within thousands of cases, and then used that information to perform other analyses critical to the accurate valuation of financial assets to be acquired by the company, resulting in greater efficiency and reduced risk and cost.✔️Created several web scrapers to aid in the retrieval of publicly available information of legal cases. ✔️ Employed AWS to run instances of cloud computing to scale up the scrapping operation and perform inference with ML models, which brought greater fluidity in the company clients screening and speed up in the information retrieval processes.

Región Metropolitana De Santiago, Chile

✔️Applied NLP (Transformers BERT) to classify text data and duplicate the information we had about the competition. Also performed unsupervised topic modeling on the same data, uncovering the main topics that customers were mentioning about the company and our competitors. ✔️Utilized DBSCAN to perform clusterization on our clients database and RNNs to predict customer satisfaction, which in turn enabled a better planification of our customer satisfaction strategies.✔️Designed data pipelines using python to reduce manual processes, for example a process that took 3 senior analysts 2 days to complete became one that took a junior analyst half a day to finish. This process was repeated twice a month. ✔️Python pandas to enable big data analyses (+10 Million rows) and present them in Power BI (using DAX, M), powering my area’s transition from excel to open-source and business intelligence software.

Santiago, Región Metropolitana De Santiago, Chile

✔️Used SQL to feed a dashboard that calculated the bonuses for the ~1000 salesmen of the company. Had to coordinate with several areas such as human resources and IT. ✔️Maintenance and optimization of SQL code. Reduced the running time of processes that feed for the company clients dashboard from 45 minutes to <1 minute. ✔️ Automated excel reports using Pandas, Numpy, SQLAlchemy and Power Query, saving hundreds of hours of work of manual excel manipulation.✔️ Supported the migration of our SQL Server database to Azure Cloud, enabling the use of cloud databases in our area.

Santiago, Región Metropolitana De Santiago, Chile

✔️Automated manual data processing using SPSS, for example a process that took a senior analyst a 1,5 days became a process that was finished on the spot (<2 minutes). ✔️Designed ETL pipelines using SPSS to extract data from large databases. ✔️ Maintained monthly reports, using excel, about company sales and goals, providing timely information to management, which was used for sales monitoring and strategic planning.