Alfonso Barajas Cervantes

---

About me

Professional Career

• Data Scientist - SAGEGIS - Manassas, VA (Remote) | Oct 2024 - Present
Reporting to CEO
Objective to reach: Enhance security resilience by improving Intrusion Detection Systems (IDS) through a robust Data Science pipeline guided by the CRISP-DM methodology, focusing on Anomaly detection. Build binary classifiers to capture adversarial behavior, build interpretable models for accurate risk assessment.
Data-Driven Successes and Implementation Outcomes

• Reduced Incident Response Time 60%, by applying faster methods on test time prediction
• Increased Anomaly Detection Accuracy by 35% by using XGBoost and CatBoost
• Benchmarked 13+ML models and hypertuned models in a K CV fashion for optimal performance
• Used SHAPley graphs to interpret model predictions
• Research Assistant - Northeastern University - Boston, MA | Feb 2023 - Dec 2024
Reporting to Profesor Pedrag Radivojac
Objective to reach: Contribute to the advancement of AI-driven variant classification in rare disease diagnostics by applying Bayesian Networks. Focused on building interpretable probabilistic models to reduce diagnostic uncertainty and enhance clinical decision-making.
Data-Driven Successes and Implementation Outcomes

• Reduced diagnostic uncertainty by 25% through interpretable Probabilistic models
• Improved accuracy in genetic variant classification by 6% by using Bayesian Categorical Networks
• Mitigated operational risk by uncovering interdependencies in evidence codes, leading to enhancements in clinical decision-making frameworks.
• ML Engineer - SynapyicSolutions.ai (Consulting Firm) -Berlin, Germany (Remote) | Oct 2024 - Nov 2024
Reporting to CEO
Objective to reach: Drive model improvement of accuracy for time series forecasting and anomaly detection by leveraging SQL pipelines and large language models (LLMs) via Vanna.AI, with a strong focus on data quality, standardization, and human-in-the-loop feedback.
Data-Driven Successes and Implementation Outcomes

• Improved answer accuracy by 30% in SQL Queries by curating and standardizing SQL documentation
• Reduced user query resolution time by 15% by guiding LLM responses via using ground truth SQL
• AI and ML Engineer - Auris Labs - Bangalore, India (Remote) | Apr 2024 - June 2024
Reporting to CTO
Objective to reach: To advance enterprise cybersecurity and operational intelligence through research-driven anomaly detection with an emphasis on guiding the application of the most suitable models based on available data, infrastructure, and business requirements.
Data-Driven Successes and Implementation Outcomes

• Documented Anomaly Detection Strategy across Statistical-based, Classification based and Nearest Neighbor-based models based on data availability, server infrastructure.
• Research Assistant - Northeastern University - Boston, MA | Sep 2022 - Feb 2023
Reporting to Profesor Saiph Savage
With 1 direct report: Graduate Data Science student
Objective to reach: To address the challenges of low-quality annotations in the AI data-labeling ecosystem by designing a recommender system for crowdworkers. The goal was to enhance the task selection process for workers in crowdsourcing platforms, improving both the quality of their work and their ability to meet personal goals.
Data-Driven Successes and Implementation Outcomes

• Developed Recommender System Framework that help crowdworkers select tasks based on goals
• Identified Key Design Implications by doing Qualitative Analysis of 4 Interviews of 1 hr each

Projects

Explainable Sentiment Analysis Of Patient Drug Reviews | PyTorch, SHAP, LIME, Generative LLM, Google Gemma | (Apr 2024)

• Used Transformers AutoTokenaizer to tokenize over 200k patient drug reviews
• Built a binary sentiment classifier using Google BERT model with pretrained encoder
• Used interpretable model explainers SHAP and LIME to obtain token contributions per class
• Finished with a summarize each drug what patients loved and complained about.

Understanding interplay of evidence codes in ACMG/AMP guidelines through Bayesian Networks | bnlearn, Scikit-learn, pgmpy | (Dec 2023)

• Understand dependencies on factors that affect pathogenicity in missense variants using Clinvar, gnomAD databases.
• Usage of Bayesian Probability such as Prior, Likelihood and Posterior to derive a scale of pathogenicity. Predicted pathogenicity of variants Via Bayesian Networks.
• Provide of a Clinical Framework for clinicians and researchers to guide decision making of de novo pathogenicity prediction

Kanbas Website App | React.js, Node.js, MongoDB, Express.js | (Apr 2023)

• Developed a full-stack, single-page web application simulating a course management system similar to Canvas
• Implemented features like user authentication, dynamic course and assignment management, and grade tracking.
• Built RESTful APIs with Node.js and Express to handle CRUD operations, and used MongoDB for data persistence.
• Frontend built with React and React Router, enabling seamless navigation and real-time data updates using Axios.

Recommender Systems for Crowdsourcing | Research Project, A/B Testing, UI/UX Design | (Dec 2022)

• Address research questions on how to address goals for crowdworkers and what challenges do they face in their work environment in order to benefit crowdworkers.
• Conduct A/B Testing via qualitative research (interviews and surveys) with crowdworkers to test features of our Add-on or Plugin.
• Specifically, testing two variants of our Recommender System following a content-based or collaborative-filtering approach.
• This guides our system design and provides detailed principles to build a Recommender System suited to the needs of the User.

Research Forecasting Finance Time Series with TDA | pytorch, giotto-tda, statsmodels, scikit-learn | (Aug 2022)

• Understand one of the state of the art algorithms to forecast time series within Cryptocurrencies.
• Developing a robust methodology to forecast using using Topological Data Analysis
• Compare different deep learning and classical architectures for time series forecasting

COVID-19 Prediction | pytorch , pmdarima ,statsmodels, TSPred(R) | (Jun 2022)

• Used LSTM architecture in UN Youth Hackathon and won a 3rd place against data scientists around the world
• Provide of scientific understanding of ML-AI methods vs Classical methods for TS Forecasting
• Predictions and Performance Metrics comparison gainst LSTM, ARIMA, and ARIMA with Kalman Filter
• Found that LSTM fits better for both Confirmed Cases and Deaths data for most countries.

Cryptocurrencies Critical Transition: Detecting it in Time Series with TDA | giotto-tda, pandas, numpy | (Dec 2021)

• Replicating results of scientific paper to detect critical transitions in Cryptocurrencies.
• Understanding the robust method of the alarm using TDA - Topological Data analysis along with K-means
• Found that using this visual alarm is efficient to detect critical transitions in volatile seasons

Stock Market and Sentiment Analysis for Investing Based in Bitcoin | Btalib, nltk, pyspark | (Jun 2021)

• Comprehend technical analysis and risk in trading topredict price direction
• Prediction along with indicators from NLP and Sentiment Analysis from large histocal dataset of tweets
• Found that serious news instead of tweets, would be a better indicator to predict direction.

E-commerce business project using CRISP methodology | Seaborn, Matplotlib, Pandas | (Jun 2021)

• Focused on data cleaning and building forecasting models of demand of products
• Make prediction on the delivery of products in relationship with dimension and price factors
• Propose techniques to handle and avoid errors on data.

Optimization Cost-Nutrition in Raw Vegan diet | Pandas, Numpy | (Feb 2021)

• Designed a Mathematical Model that provides an efficient solution.
• Knapsack 0/1 was the implemented algorithm with dynamic programming
• Given a limited budget , provided the list of fruits, vegetables and oilseeds, the output is the best combination possible that maximizes nutritional content

Polynomial Approximation and Weather Forecast | Numpy, Pandas | (Nov 2020)

• Implemented a efficient solution using Cholesky Method to give a polynomial formula of degree 3 that fits the best with the data.
• Proposed a segmentation of 3, 5, 10 year period and with seasons
• Deal with large data sets and figure out how to load it to work with.

Neo4J: PROFECO | Py2neo | (Jun 2021)

• Develop a graph database consistent with business rules
• Frontend made with Python with motor search connected to Neo4J
• Design an application that allows certain queries related to products and stores

MongoDB: NYC Bike Ride System | Pymongo,pprint | (May 2021)

• Design and develop a document database consistent with business rules
• Load data using batch method due to large amount of data.
• Propose trips that full fill parameters like time or that the user choose.

Cassandra: Book platform | Pandas, Cassandra | (May 2021)

• Design and develop a column-family database consistent with business rules.
• The books platform let the user add information about their favorite books, etc.
• The admin of the app could make queries and basic statistics on it.

Redis: Url shortener | Redis, Requests | (Apr 2021)

• Design and develop a key-value database consistent with business rules.
• System would support user management such as create new account, delete, etc.
• User could add urls to wishlist and categorize it within topics.

Achievements

I am proud to have been a member of the All-Student team that competed against over 70 teams from 33 different countries, and even more proud to have contributed significantly to our team's success. Specifically, my contribution to the project involved constructing an LSTM neural network to predict new coronavirus cases and deaths per country, which helped our team secure a well-deserved 3rd place in the UN Youth Hackathon . The following are lists of achievements:

• Winner Contetant UN Youth Hackathon (Dec 2021):

  Mission: Create solutions to fast track progress towards the UN Sustainable Development Goals (SDGs) using data.

  Certificate of Achievement: 3rd Place :3rd place out of +70 teams from 33 countries in 4 regions in the world
  Certificate of Achievement: All-student team: The winning team was composed entirely of students at the time of the contest.
  Certificate of Participation
• Winner Contestant ICPC (Oct 2021):

  The International Collegiate Programming Contest provided me of skills how to work in team to solve algorithmic problems under pressure.
  Results: Top 10% - Advance to ICPC Regionals in MX, Zacatecas 2022

  Certificate of Achievement
• Scholarship SAS LATAM 2021 (Jul 2021)

  Awarded a competitive scholarship which includes live mentorship from renowned SAS experts. The scholarship comes with a generous fund allocation of up to $9,250 USD for live mentoring on five Data Science/ML courses.

  Some courses taken: Applied Analytics using SAS Enterprise Miner; Sas Visual Statistics on SAS Viya: Interactive Model Building

• Winner of Talent Award of the University Bachelor: Research (June 2018)

After a rigorous selection process that included the most competitive high schools in Mexico, I was honored with a prestigious recognition in Scientific Research. This distinction was awarded to me for my contribution to promoting and generating scientific research in the natural sciences, Mathematics, sharing knowledge, and presenting at institutional, state, and national levels. My journey towards this accomplishment began at the age of 11 and continued until I turned 18. (UNAM News)

Certificate of Achievement
• Gold Medal/First Places in Mathematical Olympiads - State Level (2016-2018)

  (2018) Won Gold Medal in the 7th Knowledge University Olympiad in the field of Mathematics. UNAM, MX

  (2016-2017) First place in the 31st and 30th Mathematical Olympiad of Mexico City. Both participations required me to take a full preparation and rigorous exams where I went until the last final stage Pre-selection of the Mexico City team in the Mathematical National Contest. This i did it while achieving high performance in High School, being one of the best in my generation.

  (2016) Won Bronze Medal in the 6th Knowledge University Olympiad in the field of Mathematics. UNAM, MX

  Certificate of Achievement
• Within First places in Children's Knowledge Olympiad - State Level (2015)

  In order to secure a scholarship from BBVA Bancomer for my high school education, I had to take a rigorous exam in ninth grade. This exam was administered to approximately 1,000 students, who had won the Children's Knowledge Olympiad (2011-12). Out of those 1,000 students, only around 50-70 were ultimately awarded the scholarship.

  Certificate of Achievement
• 2nd place in Mathematical Olympiad - National Level (2013)
  

  I earned a silver medal at the national level in the "XIV Competencia Cotorra de Matemáticas" a mathematical contest designed for students under the age of 12.

  Certificate of Achievement
• 1st place in Children's Knowledge Olympiad Contest - State level (2012)

  As a sixth-grade student, I took multiple rigorous exams that were administered to students across the country. The number of students that took the exam were approximately 2.2 million students. I achieved the highest score in the State of Mexico, which is the most populous state out of 32 Mexico's states. As a result, I was awarded a scholarship from BBVA Bancomer to fund my secondary school education. Video

  Certificate of Achievement

Sharing Mathematical Knowledge

• CS 6140: Discrete Structures (Fall 24')

  Teaching Assistant in a Computer Science Graduate level course at Northeastern Unviersity. Help master student learn machine learning concepts during Office Hours each week during the semester. Average weekly time spend close to 20 hours. Grade students’s assignments and projects related to an innovation in Machine Learning. Topics lectured on class Probability, MAP, MLE, Optimization, K-means, Linear Regression, Logistic Regression, Regularization, Perceptron, NN, SVM, PCA

• CS 1800: Discrete Structures (Summer 24')

  Teaching Assistant in a Computer Science Undergraduate level course at Northeastern Unviersity. Help undergrad student learn mathematical and computational concepts within Computer Science Theory during Office Hours each week during the semester. Average weekly time spend close to 20 hours. Grade students’s assignments and exams related to mathematics: combinatorics, proofs and concepts that help build foundation to Computer Science. Guide them towards a successful learning.

• DS 5220: Supervised Machine Learning and Learning Theory (Fall 23')

  Teaching Assistant in a Data Science Graduate level course at Northeastern Unviersity. Help master student learn mathematical and computational concepts within Machine Learning Theory during Office Hours each week during the semester. Average weekly time spend close to 20 hours. Grade students’s assignments and projects related to machine learning. Guide them towards a successful learning experience in Machine Learning

• Training middle school students for Mathematical Olympiad (Fall 19')

  Taught combinatorics, number theory and geometry to middle school students. This was part of a session of 4 hours of training to participate in final stages on Middle School Mathematical Olympiad

  Create a series of exercises to solve during in-person session training along with the young talented students.

  News | Example of third training session
• Founded MateABC and MateABC English (2019-2020)
  

  Founded MateABC (in English as well) to promote and divulge Mathematical content to be more accesible. I created a Facebook page where I created a community up to 3,800 followers interested in Math solving skills, history of mathematics and mathematical challenges open to whoever wanted to participate from Latin American countries.
  Created 50 videos sharing Modern Geometry, Number Theory, Calculus concepts and showed ways to improve mathematical thinking through series of videos with math challenges. Both videos are in English and its version in Spanish. MateABC English (241 subscribers) and MateABC (978 subscribers)

Workshop given

• How to solve problems in math contests? (2018)
  

  I gave a workshop in my final year of high school to classmates from 9th to 12th grade. There I was able to introduce techniques in how to tackle problems in math contests, as It is one of my hard skills. I was able to encourage young people to participate more in this type of contests.
  Evidence of Participation

Workshop attended

• Critical Assessment of Genome Interpretation - (2023) - Boston, MA
  Attended workshop where leading researchers talked about the innovation and progress done and to be done towards shaping the future of computational Genome Interpretation. Keynote Speakers: Vence Bonham Jr. , National Human Genome Research Institute ,Wendy Chung, Boston Children's Hospital , Heidi Rehm, Broad Institute of MIT and Harvad

• International Masterclass: Hands on particle physics (2018) - University City, Mexico City
  During high school, I had the opportunity to participate in an introduction to research program on the fundamentals of matter and forces at a particle sciences research center affiliated with UNAM in Mexico. Performed experiments on real data from real experiments at CERN. I was able to participate in an online discussion with researchers in the field of particle physics about the results from the experiment.

  Certificate of Participation

Education

2022-2024	Master of Science in Computer Science at Northeastern University Relevant Coursework: Machine Learning, NLP, Algorithms, Research Methods in Human Computer Interaction, Web Development
2018-2022	BSc in Data Science-Mathematics at National Autonomous University of Mexico Relevant Coursework: Deep Learning, Machine Learning, NLP, Time Series, Data Mining, Probability and Statistics, Relational and NoSQL Databases, Discrete Mathematics, Big Data I and II, Calculus 1-IV, Linear Algebra I-II,
2015-2018	National Preparatory School 9 "Pedro de Alba"

LinkedIn : Alfonso Barajas

Github : AlfonsBC

YouTube : MateABC English

YouTube : MateABC

Facebook : Facebook Page MateABC