Evelyn Cusi Lopez

Hello I'm

Data scientist / Software engineer

Download CV

About Me

Hi, I'm Evelyn, a Data Scientist Team Lead at Datec LATAM, passionate about turning advanced AI into scalable, high-impact solutions.

At Datec, I lead the design and implementation of intelligent agents, RAG systems, and predictive models, while mentoring junior talent and ensuring business needs are translated into robust technical solutions. My role allows me to combine leadership with hands-on work, creating innovation that truly drives results.

My passion for technology started early in university, where I explored applied AI and systems engineering through my theses. Later, I joined Google Summer of Code (2019) with the Pharo Consortium, a pure software engineering project that strengthened my development skills and gave me my first experience mentoring within a global open-source community.

To deepen my expertise, I pursued a Master’s in Data Science at the Pontificia Universidad Católica de Chile, graduating with distinction after earning the highest grade. This milestone reaffirmed my commitment to excellence and my drive to push the boundaries of applied AI.

Today, I continue to blend my background in software engineering and data science to build AI solutions that are innovative, reliable, and impactful.

Experience

Data Scientist Team Lead - Datec LATAM - D4G (Data for Growth) (Jan 2025 – Present)

  • Lead the AI and Data Science business line, designing end-to-end architectures and strategies for internal and client projects.
    • Coordinate multidisciplinary teams, mentor junior talent, and act as a technical point of contact for strategic clients.
    • Design intelligent agents and RAG solutions with OpenAI, LangChain, and Pinecone, transforming chatbots into secure and autonomous systems deployed on Google Cloud.
    • Direct high-impact projects, including:
      • Rolling Forecast and purchase automation (500+ SKUs) in Databricks using SARIMAX, reducing planning time and human error.
      • Optimization of university resource allocation through predictive models and business rules.
      • Document extraction with OCR (Llamaparse), Generative AI (OpenAI), and Computer Vision (YOLOx, Roboflow), structuring unorganized information with LLM support.
    • Provide training and mentorship to junior Data Scientists; conduct technical interviews to support hiring and team growth.

Data Scientist - Datec LATAM - D4G (Data for Growth) (Dec 2023 – Dec 2024)

  • Executed advanced analytics projects and AI prototypes for e-commerce and external clients, focusing on business impact and process optimization.
    • Built a purchase propensity model for marketing, prioritizing the identification of false positives to enable more effective campaigns and higher conversion.
    • Designed and implemented an automated product onboarding system: Generative AI (OpenAI) generated descriptions from EAN codes, and web scraping retrieved images, enabling catalog uploads in minutes instead of hours.
    • Launched the first version of an internal documentation RAG with LangChain and Pinecone, improving information retrieval efficiency; mentored a junior engineer to ensure continuity and quality.
    • Supported client consulting, participating in needs assessments and drafting strategic proposals with the Head Lead.

Data Scientist - project44 (August 2021 – Current)

  • Build predictive models using machine learning tools to predict ETA and ETD (Estimate Time of Arrival and Estimate Time of Departure).
    • Data cleaning and preparation:
      • Design algorithms/heuristics to detect anomalies in tracking data: GPS signals, ETA and ETD.
      • Data visualization, the main tools used were: matplotlib, seaborn, folium and kepler.
      • Data augmentation.
    • Train and evaluate the model, mainly used algorithms were CatBoost and XGBoost.
    • Develop automatic reports of model’s performance, using AWS and Jenkins.
  • Improve data quality for a p44’s own reverse geocoding system.
    • Collect external and internal data (using snowflake/SQL).
    • Analyze data.
  • Perform analysis based on historical data, to improve hyperparameters in pre-existing ML systems.

Mentor - Google Summer of Code 2021 - Pharo Consortium (May 2021 – July 2021)

Mentoring of the Refining Code Critics project of Alejandra Cossio student, for Pharo Consortium

Software Engineer - SEMANTICS S.R.L. (August 2018 – July 2021)

  • Develop web page for registration to ESUG Conference 2019.
  • Train junior engineers.
  • Develop, improve, manage and support SmallSuiteGenerator (Smalltalk tool to generate test cases automatically using Genetic Algorithms).
  • Develop and improve Refactoring tools for Pharo consorsium.

Software Engineer - Google Summer of Code - Pharo Consortium (May 2019 – August 2019)

I made contributions to Pharo 8, among the main contributions are:

  • Fix refactoring issues.
  • Improve existing refactorings.
  • Add new refactorings.

Summary website: https://lin777.github.io/gsoc2019/

Achievements

Predicción del bajo rendimiento académico en ámbitos rurales de Bolivia mediante modelos de aprendizaje automático (2025) Jornadas de Jóvenes Investigadores AUGM

This study analyzes academic performance in a rural school in Santa Cruz, Bolivia, using student records from 2015 to 2024. The analysis revealed high failure rates in key subjects such as Natural Sciences, Mathematics, and Language, representing nearly half of all cases. To anticipate failures, several supervised machine learning models were trained. Among them, CatBoost achieved the best performance with an F1-score of 0.84 and 91% accuracy, surpassing models like XGBoost, Random Forest, SVM, and neural networks. The model enabled predictive lists of at-risk students for the coming years, supporting teachers with early and targeted interventions.

Find out more

TestEvoViz: visualizing genetically-based test coverage evolution (December 2022) Empirical Software Engineering

This paper presents TestEvoViz, a visual technique to introspect genetic algorithm-based test generation processes. TestEvoViz offers the practitioners a visual support to expose the process and decisions made by the generation algorithm. We first present a number of case studies to illustrate the expressiveness of TestEvoViz...

Find out more

TestEvoViz: Visual Introspection for Genetically-Based Test Coverage Evolution (September 2020) VISSOFT

Genetic algorithms are an efficient mechanism to generate unit tests. Automatically generated unit tests are known to be an important asset to identify software defects and define oracles. However, configuring the test generation is a tedious activity for a practitioner due to the inherent difficulty to adequately tuning the generation process...

Find out more

Visual automatic fire detection through the Histogram of Oriented Gradients and Support Vector Machine (October 2016)Scielo

The visual detection of objects in digital images, is one ofthe challenging issues in Computer Science, especially ifthose objects do not have a definite shape, as in the case offire. To meet the challenge, it is necessary to use techniques Computer Vision and Machine Learning, so that the results are approximate as much as possible, to the human visual system. Given a digital image, with the aim of identifying areas where fire there, the computer should locate exactly the same areas as a human would...

Find out more

Latest Projects

air quality project

Integration of Machine Learning into Chile’s National Air-Quality Forecasting System

This project integrates advanced Machine Learning and Deep Learning techniques into Chile’s official WRF-MMA air-quality forecasting model to improve PM2.5 prediction. Using models such as XGBoost, LSTM, and CNN-LSTM, the system reduced forecasting errors, improved sensitivity to critical events, and avoided false alarms. The solution is being integrated into the national system, modernizing Chile’s forecasting capacity with greater accuracy, adaptability, and operational value for environmental management.

Find out more

project name

Classification of native languages of Bolivia

The objective of this project is to create a language classification tool, which, although it is a simple task, is elementary to carry out more complex tasks such as automatic translation, sentiment analysis, speech to text, text to speech conversion, etc.

Find out more

project name

Python Roadmap

The purpose of this project is to trace a learning path from scratch to become an AI expert, with this I hope to help many others to take the first step in this fascinating area.

Find out more

project name

Driving Anomaly Detection

This paper describes the development of a mechanism to detect driving anomalies, which is implemented using a mobile device and Machine Learning techniques. The objective is create a tool capable to find anomalous behaviors in the driving of human or autonomous agent...

Find out more