https://github.com/daniellevenstein/daniellevenstein
Data Engineer & AI Developer showcasing samples of work.
https://github.com/daniellevenstein/daniellevenstein
data-science image-classification machine-learning
Last synced: about 7 hours ago
JSON representation
Data Engineer & AI Developer showcasing samples of work.
- Host: GitHub
- URL: https://github.com/daniellevenstein/daniellevenstein
- Owner: DanielLevenstein
- Created: 2026-02-24T20:39:19.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2026-06-21T17:21:30.000Z (21 days ago)
- Last Synced: 2026-06-21T19:03:25.546Z (21 days ago)
- Topics: data-science, image-classification, machine-learning
- Homepage: https://www.kaggle.com/code/daniellevenstein
- Size: 3.96 MB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Data Engineer & Cloud Developer
Software engineer with 6+ years of experience working in AWS environments and data systems. Recently focused on machine learning and data analysis projects, including clustering, and business analytics.
🔗 [Deployed Models](https://huggingface.co/DanielLevenstein)
🔗 [Kaggle Notebooks](https://www.kaggle.com/daniellevenstein)
## Professional Work:
- Developed Java/Scala scripts to migrate and normalize legacy financial records into updated schema formats for downstream validation workflows. (FINRA)
- Built a Lambda-based system to track application uptime and system health signals. (b.well)
## Certifications:
- Certified AWS Cloud Practitioner (AWS, 2025)
- Post-Graduate Certificate in Artificial Intelligence & Machine Learning (UT, 2026)
## Featured Projects:
*Figure: Shows an image of my AWS Rag and AWS Certification coach project side by side.*
## Detailed Information:
🎓 AWS Certification Coach (ML Evaluation, Streamlit, Docker)
🔗 GitHub: [DaielLevenstein/AWS-Certification-Coach](https://github.com/DanielLevenstein/AWS-Certification-Coach)
🔗 Live Demo: [aws-certification-coach-onrender](https://aws-certification-coach-latest.onrender.com/)
**Overview:** Developed an AI-powered study application for AWS certification practice. The app presents freeform AWS exam-style questions, evaluates learner answers with a trained local classifier, and returns structured coaching feedback with score, missing concepts, detailed answer guidance, and original multiple-choice provenance.
**Key Features:**
- Built a Streamlit study interface for answering AWS certification questions in freeform text.
- Generated self-authored AWS exam-style training, holdout, and sample question datasets.
- Designed a combined JSON artifact format that stores questions, original multiple-choice provenance, correct answers, wrong answers, and partial-credit examples together.
- Trained a local answer classifier to evaluate full-answer correctness with release metrics for accuracy, precision, recall, and confusion matrix results.
- Created a partial-credit regression model using continuous answer ratings and mean-squared error evaluation.
- Added validation tests for classifier behavior, low-credit answer rejection, memory overhead, generated artifacts, and model performance.
- Containerized the application with Docker for local deployment and Render-ready hosting.
**Tools & Technologies:** Python, Streamlit, Docker, Machine Learning, Classification Models, Regression Models, JSON Data Pipelines, Automated Testing, AWS Certification Content
---
### 🎺 AWS Documentation RAG System (LLM, Vector Search, ETL Pipeline)
🔗 [GitHub: Aws-Documentation-Rag](https://github.com/DanielLevenstein/aws-documentation-rag)
**Overview:**
Developed a retrieval-augmented generation (RAG) application for querying AWS documentation. Built a pipeline for crawling AWS documentation pages, preprocessing and storing data in a vector database, and generating technical responses using a language model through a Streamlit interface.
**Key Features:**
- Developed a web crawler for ingesting AWS documentation pages.
- Kept track of feature downloads and crawl depth to prevent redownloading files repeatedly.
- Implemented vector-based semantic retrieval for documentation search workflows.
- Integrated a language model for context-aware technical question answering.
- Built a Streamlit frontend for interactive querying and response generation.
- Containerized the application using Docker for local deployment and testing.
Tools & Technologies: Python, Streamlit, Docker, Vector Databases, LLMs, Web Scraping, ETL Pipelines
---
### 🪖 Helmet Classification Pipeline (CNN, Training Data Curation)
🔗 [Interactive Streamlit App](https://huggingface.co/spaces/DanielLevenstein/Helmet_CNN_Data_Quality_Case_Study)
🔗 [Helmet CNN Training Notebooks](https://github.com/DanielLevenstein/Helmet_CNN_Model_Training_Notebooks)
**Overview:**
Developed an image classification pipeline for helmet detection using a CNN trained on standardized 100x100 image inputs. Build a data curation workflow to transform raw annotation images into standardized size and remove samples with too low a resolution to be useful.
**Key Feature:**
- Built four CNN-based image classification models using four different training datasets.
- Evaluated models based on a holdout dataset consisted of equal numbers of samples from Dataset2, and Dataset3.
- Improved accuracy of a classification model from 77% to 94%, and calculated precision, recal, and f1 metrics for each model.
- Identified categories of images model struggles with for further fine-tuning.
Tools & Technologies: TensorFlow, Keras, OpenCV, Streamlit, HuggingFace
---
### 🍕 FoodHub – Delivery Data Analysis
🔗 [GitHub: FoodHub Data Analysis](https://github.com/DanielLevenstein/FoodHub_UT_ML_Project1)
**Overview:**
Developed discount program for a food delivery app, which reaches up to 12% of the population under model conditions.
**Key Features:**
- Cleaned dataset and imputed missing ratings.
- Identified top-rated restaurants by cuisine.
- Calculated discount cost vs. total customer reach.
*Figure: Discount cost vs total reach with value per customer analysis.*
---
### 💪 Stress vs. Physical Activity – Kaggle Data Analysis
🔗 [Kaggle: Stress vs Physical Activity Correlation](https://www.kaggle.com/code/daniellevenstein/stress-vs-physical-activity-correlation)
**Overview:**
Analyzed relationships between stress levels and physical activity using real-world survey data.
**Key Features:**
- Performed exploratory data analysis and correlation studies
- Generated three population clusters using K-Means
- Identified occupation-based stress correlations
- Visualized cluster separation and activity distribution
*Figure: K-means clustering reveals distinct population-level groups based on total activity and stress.*
## 📫 Contact
Phone: 512-885-5925
LinkedIn: https://www.linkedin.com/in/daniel-aaron-levenstein/