Spanish version here

Freelance computer engineer with 15 years of experience. Specialized in data projects: data development and cloud infrastructure. Proficiency in Python for backend development, data analytics, machine learning and AI processes. Working as an individual contributor or with my team (skilled freelance data scientists and web devs). CTO and Tech Lead in several startups and projects.

☎️ Contact

📧 Email: luzfrias at circiter dot es

🐦 Twitter: koldLight

🐙🐱 Github: koldLight

🔗 LinkedIn: Luz Frias

🔗 Company website: Circiter

👩🏻‍💻 Projects

Marketing data AI chatbot

Roya - USA marketing startup

Designed and built an AI-powered pipeline to summarize marketing data and provide an interactive chatbot for business insights. The chatbot enables users to ask questions about sales trends, conversion performance, and top- and bottom-performing products.

Technologies: OpenAI platform integrated in R and Shiny.

Brain Signal Recording and Data Analysis

Decedario – Spanish startup

Recording of EEG signals using an OpenBCI signal acquisition device, along with the software needed to log sessions and provide an analytics dashboard for neuropsychology professionals. The system is designed to evaluate sessions conducted by professionals with patients with neurodivergences or brain injuries.

I supported the startup throughout the process of selecting the device, applying to accelerator programs, building PoCs, and developing and deploying the final system.

Technologies: Python, OpenAI, Streamlit, MNE.

Geospatial System and Viewer for Rural Land Parcels

Hareas – Spanish startup

Responsible for the backend, database, and infrastructure of a system providing enriched information on rural land parcels in Spain. The platform is designed for land purchase and sale professionals, farmers, and other stakeholders. Serverless infrastructure design to handle computing and traffic peaks and reduce costs.

Technologies: AWS, Lambda, Martin, Python.

Vehicle Rental Search Engine

Oscar Car Rental - Dutch startup

Complete replacement of the product’s vehicle rental search system. Achievements:

• Average speed improvement of 400x, from several seconds to under 100 ms.
• Significantly lower database load, resulting in reduced costs.
• New features that improved conversion.

Technologies: Python, Polars, Kubernetes.

Mortality Monitoring (MOMO)

Instituto de Salud Carlos III (ISCIII) - Spanish public health organization

Automation of the process of mortality data ingestion from civil registries, processing, modeling and generating reports and alerts from MOMO. The objectives were to alert and measure excesses and attribute mortality to events such as heat waves or flu.

MOMO received a lot of attention during the Coronavirus health crisis and it was a key project to estimate the real impact of the epidemic on Spanish mortality.

🔗 MOMO

🔗 Kairos

During the health crisis, I also developed the ISCIII data publication and monitoring panel, and the calculation of the reproducibility index by region, which was reported to other organizations.

🔗 COVID-19 Dashboard

Main technologies: R, Python, Docker, Flask, Flexdashboard.

Modeling techniques: GAM, mixed-effects models, survival analysis.

Sentinel surveillance system on respiratory infections

Instituto de Salud Carlos III (ISCIII) - Spanish public health organization

Design and build the new surveillance system for the endemic phase of the COVID-19, and the rest of existing respiratory infections such as the influenza virus. The main objective of this application is to unify the data provided from the different autonomous communities of Spain and run automatic rules and validations to ensure data integrity. This system also enables the administrators from changing the required data without code upgrades, so it adapts quickly to regulation changes.

🔗 System information (spanish only)

Main technologies: Python, Docker, Flask, MySQL.

Marketing analytics and optimization dashboard

Roya - USA marketing startup

Create a demo platform for the product. The client was a NY based startup with a strong analytical product for optimizing marketing processes in big companies. This included:

• Code refactorization, from a local version to a production one for a R Shiny dashboard
• Re-styling to switch to a more modern interface, matching the company guidelines
• Wrap everything behind a user-control
• Server Setup from scratch using Digital Ocean cloud provider

Main technologies: R, Shiny, Shiny Server, flexdashboard, Python, Flask, Linux, nginx.

Preventive maintenance on wind generators

Atlantica - Renewable energy

The main objective was the development of a heating alarm system on the electric transformers installed in multiple wind plants in South-America. The project included a detailed Exploratory Data Analysis, data inconsistency detection, and the heating model. The biggest challenge was the feature engineering, to take into account the physical rules and the heat behavior over time.

Main technologies: Python, Jupyter Notebooks.

Modeling techniques: GAM.

Transit analysis in physical stores

Johnson Controls - Buildings infrastructure and IoT

Development of a system to analyze people transit in physical spaces through cameras. My main objectives were:

• Set up the infrastructure, the store management websites and the cameras and the transit analysis dashboard.
• Create the heatmaps on the store plans based on the information received from the cameras.
• Design and develop a dashboard to analyze the hot and cold areas, the user behavior and the main KPIs across the stores.

Main technologies: R Shiny, Python Flask, Bootstrap, Google Cloud (Compute Engine, Cloud Functions and BigQuery).

Modeling techniques: trigonometry, minimization of error functions.

Data Warehouse and dashboard for e-scooter sharing service

Acciona Motosharing - Mobility

Design, develop and deploy the Data Warehouse to analyze the e-scooter sharing quality service, follow the vehicles incidences, the customers behavior and take strategic decisions. I led the tech team.

The goals we achieved:

• Connect different data sources, such as the application backend and the bike log system in BigQuery, apply the business logic and register all the main KPIs in the DWH.
• Automatize all the ingestion processes involved in a robust pipeline in Google Cloud Platform.
• Design and develop an interactive dashboard for different kind of users (Quality, Business and Marketing teams), including data restriction access depending on the role.
• Mentor and train the client’s internal team to maintain and evolve the system.

Main technologies: Python, Flask, Dash, PostgreSQL, GCP.

Interaction prediction in email marketing

MDirector - Email marketing platform

Development of a prediction system for openings and clicks in an email marketing system.

The objectives:

• Avoid sending to users with almost zero probabilities of clicks and opens
• Increase the reputation of the system with the email provider
• Optimize the income of the daily schedule automatically assigning emails to users

Main technologies: Python, Docker, AWS, MySQL

Modeling techniques: random forests, linear optimization.

Google Ads performance tuning

OrbitalAds - SEM optimization startup

Creation of machine learning processes to increase the performance of Google Ads campaigns. Specialization in:

• Natural language processing techniques
• Data processing and modeling
• Cloud architecture

Main technologies: Python, Docker, Google Cloud (Compute Engine, Cloud Build, BigQuery, GKE, datastore).

Modeling techniques: TFIDF, cosine distance and other NLP and other common techniques in NLP.

IoT systems management application

Johnson Controls - IoT for big corporations

Development of a management application for large IoT systems in buildings to measure supplies use, temperature, air quality and others. Integrated in Metasys, a world’s leading building automation system.

🔗 Metasys

Main technologies: Python, Flask, React.js, PostgreSQL.

Industry production planning

Reny Picot - Dairy industry

The objective was to propose quantities to be produced per product to minimize losses due to stock breakage or expiration. We develop a demand forecasting model, a stock evolution simulator and a web panel to view and download the results.

Main technologies: R, Shiny.

Modeling techniques: simulation and optimization techniques.

Renewable energy production modeling

Kren4 - Renewable-energy power plants management platform

Modeling of production in wind and solar generators based on sensor readings to detect machine failures. Integration with the customer IoT system and report automation. Development of an API to order actions and view system results.

Main technologies: ElasticSearch, Python, AWS, MySQL.

Modeling techniques: random forests.

R and Python for data analysis teaching

EAE Business School, EDIX, Bank of Spain, Acciona and other companies

Associate professor of the Big Data master at EAE. I also usually teach programming training in R and Python oriented to data analysis in other companies.

🛠 Tech skills

This is the tool set I'm used to work with:

SQL

Complex queries and optimization of these in relational databases (MySQL, Oracle, Teradata, ...). PostgreSQL at a deeper level, such as optimization of index types, DB parameters, or use of PostGIS.

Sporadic use of non-relational databases like Mongo.

Data analysis and modeling

Python (pandas, scikit-learn, numpy, ...) and R.