Resume
## Pablo LasoI am a data scientist and machine learning engineer in Nashville, TN mostly using NLP and machine learning to solve business problems and deliver efficient, scalable solutions.Pablo Laso¶
I am a Software developer, Data Scientist, Machine Learning, and Deep Learning engineer.
Master's thesis student at Harvard (Boston, MA) on AI. Also graduated from University of Twente (NL) and KTH (Sweden) in CS.
Experience at General Electric, Karolinska Institute, and recently at MICSI (start-up internship in NY, USA).
## Education<br><b>Georgia Institute of Technology</b><br/>Currently pursuing a Master's degree in computer science with a specialization in machine learning. <br/>Expected graduation: 2020<b>Belmont University</b><br/>Graduated: May 2015 (economics and statistics)Education 🎓¶
Harvard Medical School (Boston, MA, USA)
Master's thesis on AI (Deep Learning, data augmentation and generative models).
University of Twente (Netherlands)
Graduated: 2023 - M.Sc. Computer Science (specialization in Data Science).
KTH - Royal Institute of Technology (Stockholm, Sweden)
Graduated: 2022 - M.Sc. Data Science (minor in innovation and entrepreneurship).
Rey Juan Carlos University (Madrid, Spain)
[& Erasmus student exchange in Italy during 5 months]
Graduated: 2021 - B.Sc. Biomedical Engineering (specialized in Medical Imaging)xxxxxxxxxx## Experience<br/><b>The General Automobile Insurance Services, Inc.</b> | May 2017 - Present<br/><i>Data Scientist II | Sep 2018 - Present</i>• Utilize paragraph vectors (using gensim's doc2vec implementation) to perform concept detection to search for abstract concepts such as "Spanish speaking", "fracture", and "attorney representation" (among others) in claim notes • Implement online learning in order to update machine learning models in real time with new information<i>Data Scientist I | May 2017 - Sep 2018</i>• Build a nation-wide price elasticity model using a Mixed Effects/Hierarchical Logistic Regression algorithm (with cross classified and nested random effects) to predict demand for a product at various pricing structures</br>• Work closely with the AVP of Analytics and the Data Science Manager to build a model to optimize ad expense allocation for each state using price elasticity</br>• Optimize claim routing using a Support Vector Machine model (Python/SciKit-Learn)</br>• Architect and implement RESTful APIs (Python/Flask, SQLAlchemy, Celery, Redis, RabbitMQ) that expose machine learning models for consumption</br>• Perform various natural language processing tasks (R/tidytext, Python/Gensim), including sentiment analysis, topic modeling (LDA, hLDA), and training Word2Vec and Doc2Vec models</br>• Build a topic modeling dashboard for the Digital Experience team to better understand what our clients are chatting about</br>• Use topic modeling to predict if a claim needs intervention by a supervisor<br/><b>Perception Health</b> | Jan 2015 - April 2017<br/><i>Software Engineer/Data Scientist</i>• Built machine learning models (Python) that used lab data to predict if a patient will be admitted or not within a 30-day window with 75% accuracy</br>• Worked alongside Perception Health's Chief Data Scientist to analyze physicians’ referral patterns and consult hospitals on how to best reduce out-of-network referrals</br>• Used ArcGIS and ESRI data (with PostGIS) to visualize healthcare trends geographically</br>• Lead production on Perception Health's flagship product since March 2015 to the present</br>• Built customizable, interactive dashboards (Leaflet.js, D3.js, Crossfilter.js) </br>• Optimized SQL queries and data preparation for faster graph rendering and page-load times</br>• Utilized Agile methodologies (JIRA) to organize releases while ramping up features</br>• Worked with the lead architect to optimize SQL scripts across the primary product's platform</br>• Converted the platform to a RESTful API for more efficiency and to decouple the server-side code from the client side templatesExperience 👨🏼💻¶
MICSI | Jan 2024 - April 2024 | NY, NYC.
Platform and AI engineer | full time
• Developed and implemented advanced image processing algorithms, including AI-powered solutions, resulting in a 40% increase in server efficiency and reduced processing time by 50%.
• Collaborated with cross-functional teams to integrate image processing algorithms into the server architecture, leading to a 30% improvement in overall system performance.
• Optimized existing image processing algorithms through continuous testing and refinement, resulting in a 15% reduction in server downtime and improved user experience.
Martinos Center for Medical Imaging | May 2023 - Dec 2023 | Boston, MA.
ML engineer | full time
• Designed and implemented a Deep Learning model [open-source] for brain MRI WMH segmentation.
• Boosted performance by 17% by:
-. mathematically refining the loss function,
-. upgrading data augmentation methods and generative model, and
-. reinforcing the model by post-processing techniques (ensembling and uncertainty estimation).
• Implemented post-processing techniques such as ensembling and uncertainty estimation to reinforce the model's predictions, resulting in a 15% reduction in false positives.
Karolinska Institutet | Oct 2021 - Sep 2022 | Stockholm, Sweden.
Engineer Assistant | part time
• Developed and deployed a cutting-edge server system to streamline image processing tasks, resulting in a 40% increase in efficiency across all projects.
• Utilized advanced algorithms and automation techniques to reduce processing time for image tasks by 75%, improving overall project turnaround time significantly.
General Electric | Nov 2021 - Jul 2022 | Madrid, Spain.
• Developed and fine-tuned an algorithm utilizing deep learning techniques to accurately detect potential cancer areas in prostate MRI images, resulting in a 95% accuracy rate in identifying cancerous regions.
• Implemented a streamlined process to produce medical reports with detailed statistical analyses of patient data and main variables, reducing report generation time by 60%.
• Collaborated with a team of radiologists to validate the algorithm's effectiveness by testing on a sample size of 500 MRI images, demonstrating a significant increase in early detection rates by 40%.
xxxxxxxxxx## Skills<br/><b>The General Automobile Insurance Services, Inc.</b> | May 2017 - Present<br/><i>Data Scientist II | Sep 2018 - Present</i>• Utilize paragraph vectors (using gensim's doc2vec implementation) to perform concept detection to search for abstract concepts such as "Spanish speaking", "fracture", and "attorney representation" (among others) in claim notes • Implement online learning in order to update machine learning models in real time with new information<i>Data Scientist I | May 2017 - Sep 2018</i>• Build a nation-wide price elasticity model using a Mixed Effects/Hierarchical Logistic Regression algorithm (with cross classified and nested random effects) to predict demand for a product at various pricing structures</br>• Work closely with the AVP of Analytics and the Data Science Manager to build a model to optimize ad expense allocation for each state using price elasticity</br>• Optimize claim routing using a Support Vector Machine model (Python/SciKit-Learn)</br>• Architect and implement RESTful APIs (Python/Flask, SQLAlchemy, Celery, Redis, RabbitMQ) that expose machine learning models for consumption</br>• Perform various natural language processing tasks (R/tidytext, Python/Gensim), including sentiment analysis, topic modeling (LDA, hLDA), and training Word2Vec and Doc2Vec models</br>• Build a topic modeling dashboard for the Digital Skills team to better understand what our clients are chatting about</br>• Use topic modeling to predict if a claim needs intervention by a supervisor<br/><b>Perception Health</b> | Jan 2015 - April 2017<br/><i>Software Engineer/Data Scientist</i>• Built machine learning models (Python) that used lab data to predict if a patient will be admitted or not within a 30-day window with 75% accuracy</br>• Worked alongside Perception Health's Chief Data Scientist to analyze physicians’ referral patterns and consult hospitals on how to best reduce out-of-network referrals</br>• Used ArcGIS and ESRI data (with PostGIS) to visualize healthcare trends geographically</br>• Lead production on Perception Health's flagship product since March 2015 to the present</br>• Built customizable, interactive dashboards (Leaflet.js, D3.js, Crossfilter.js) </br>• Optimized SQL queries and data preparation for faster graph rendering and page-load times</br>• Utilized Agile methodologies (JIRA) to organize releases while ramping up features</br>• Worked with the lead architect to optimize SQL scripts across the primary product's platform</br>• Converted the platform to a RESTful API for more efficiency and to decouple the server-side code from the client side templatesSkills 🔨¶
- Deep Learning
- • Neural Networks, Convolutional Networks, Transfer Learning, TensorFlow, PyTorch, Keras.
- Machine Learning
- • Predictive Modeling, Classification, Regression, Unsupervised Learning, scikit-learn, NLTK.
- Programming and Development Tools
- • Python, Java, R, MATLAB, C++, SQL, Visual Studio Code, Jupyter Notebooks, RStudio, LaTeX, Markdown.
- Data Analysis
- • Data Visualization, Statistical and Exploratory Data Analysis, Seaborn, Matplotlib, Tableau, Power BI, VBA.
- Data Management
- • Pandas, NumPy, SciPy, SQL, Hadoop, Spark, AWS Redshift, DynamoDB.
- Big Data
- • Hadoop, Spark, BigQuery, Data Warehousing, MySQL.
- Business & Communication
- • Project Management, Strategic Communication, Data-driven Decision Making.
- Cloud & Deployment
- • AWS (EC2, S3), Azure, GCP, BigQuery, Model Deployment (Flask, Django, Streamlit, FastAPI).
- Version Control & CI/CD
- • Git, Docker, Kubernetes, Jenkins, Travis CI, GitLab CI.
- Languages
- • English, Spanish, French.
## Projects<br/>• Creator of SQLCell, an open source SQL client for Jupyter Notebook, which I presented at the first annual JupyterCon 2017<br/>• Creator of Achoo, a predictive system that uses allergen, air quality, and weather data to notify my son’s school nurse when he’ll need his inhaler, and was presented at AnacondaCON 2018<br/>• Contributor to <a href="https://learndatasci.com">learndatasci.com</a>, a data science blog <br/>• Creator of and primary contributor to <a href="https://lasopablo.github.io">Grid Searched</a>, a machine learning blogProjects 🙌🏼¶
• Designed a AI-based algorithm (CAD system) for Prostate MRI detection and classification. Project at Quirón Hospital, funded by General Electric (used as my bachelor's thesis) (2021).
• Contributed to the replication of a Tabletop MRI (Boston, MA, USA). Worked at LAIMBIO in Madrid, in collaboration with the Martinos Center (MGH, Harvard Medical School) (2019-20).
• Built two Deep Learning-based Lung Ultrasound (LUS) Image Recognition Systems (for Pneumonia, and COVID-19). Designed CNNs architecture and utilized Transfer Learning. Completed an internship at Alcorcón HUFA Hospital, Madrid (2020).
• Created Gest2talk, a myo-armband for people with speech impairments. Completed the project at the University of Salamanca (2019).
## Achievements<br/>• Dean’s list at Belmont University Fall 2012, Spring 2013</br>• 1 of 100 engineers nationally accepted to attend viSFest, a D3.js conference</br>• My project, SQLCell, was accepted to present at JupyterCon 2017</br>• My project, Achoo, was accepted to present at AnacondaCON 2018Achievements 🏆¶
• Harvard-MIT / HST private grant (from a Multiple Sclerosis fundation) to conduct my MSc thesis in Boston with Harvard Medical School.
• Spanish FMM "Excelencia" scholarship, among top-20 STEM students (2022).
• Swedish CSN scholarship (2021).
• 3rd prize Makeathon 4.0 (Gest2talk project), University of Salamanca (2019).
• Madrid Region "excellence" highschool diploma (2017).