CQ
Christian Quidel Data Scientist Trainee
Technical Portfolio

Featured Machine Learning Projects

End-to-end data science solutions: NLP, inference latency optimization, and containerized cloud deployment.

Lead Data Scientist & Deployment Lead

FinanceAI – Intelligent Financial Diagnostic & NLP Platform

August 2026

End-to-end Machine Learning platform engineered during the Oracle Next Education (ONE) Hackathon. Features an NLP transaction classification pipeline, a multivariate financial health diagnostic model, and multi-container cloud deployment.

Evaluation Metric
Weighted F1
Dataset Scope
240k+ Trans.
Inference Latency
< 0.8 ms (ONNX)
Architecture
4 Containers

Key Technical Highlights:

  • Categorized 240,000+ financial records across 10 classes using Scikit-learn (TF-IDF + Linear Classifier) evaluated via Macro & Weighted F1-Scores.
  • Serialized Scikit-learn pipelines to ONNX (skl2onnx) for native in-memory execution inside a Java 17 / Spring Boot backend (< 0.8ms latency).
  • Enforced strict data leakage prevention via temporal splits and optimized hyperparameters using GridSearchCV across 36 candidates (Stratified CV).
  • Orchestrated 4 microservices (Vue 3, Spring Boot, MySQL, Jupyter) using Docker Compose and deployed to Oracle Cloud Infrastructure (OCI Compute).
Python Scikit-Learn NLP (TF-IDF) ONNX Java Spring Boot Docker Compose MySQL Oracle Cloud (OCI)
Independent Data Practitioner

Predictive Modeling & Tabular Data Analytics Engine

2026

Modular experimentation environment for large tabular datasets. Implements IQR outlier filtering, continuous feature transformation, and rigorous trade-off analysis across classification and regression algorithms.

Cross Validation
5-Fold CV
Outlier Detection
IQR Filter
Evaluation Metrics
ROC-AUC / F1

Key Technical Highlights:

  • Multidimensional exploratory data analysis and statistical visualization using Pandas, Matplotlib, and Seaborn.
  • Automated preprocessing pipelines: missing value imputation, StandardScaler normalization, and One-Hot Encoding.
  • Evaluated baseline models (Logistic Regression, Decision Trees) using ROC-AUC curves, Confusion Matrices, and weighted F1-Scores.
  • Clean, reproducible technical documentation adhering to PEP 8 code standards.
Python Pandas NumPy Scikit-Learn Matplotlib Seaborn MySQL Jupyter Lab