Hi, my name is Tariere! I’m a Data Science Master’s candidate passionate about the end-to-end lifecycle of machine learning, from exploratory analysis and model selection to cloud deployment, with an emphasis on making deliberate, well-reasoned architectural decisions. I am constantly experimenting with new architectures and refining my approach to building impactful AI.
Outside of work, I’m a cozy gamer (I’ve recently started making my own mods for the Sims), a sci-fi and anime fan (Cowboy Bebop is untouchable), and someone who genuinely loves learning for its own sake. Geography and anthropology are two of my deepest special interests. I’m also an artist, I love painting, reading good books and trying out new foods…one of my favorite things about living in NYC area is that good food is always close by!
Tech Stack
Featured Projects
Economic Time-Series Forecasting on AWS SageMaker
I built this project to explore how different machine learning architectures handle the nuances of U.S. macroeconomic data. After benchmarking classical statistical models against modern gradient boosting, I found that XGBoost, paired with a 12-month lookback window and stationarity transformations, far outperformed traditional methods like ARIMA. A major part of the journey was the decision to pivot away from Deep Learning (LSTM) when it became clear the architecture wasn’t suited for the specific scale of this dataset.
Key Skills/Tech: Python, XGBoost, ARIMA, Prophet, SageMaker AI, Cloud Deployment, MLOps
RAG-Based Research Assistant
This project explores how Retrieval-Augmented Generation can make domain-specific medical research easier to work with. It ingests peer-reviewed papers on brain tumor detection, indexes them using semantic embeddings, and returns citation-backed answers through a FastAPI backend and a Streamlit interface.
Key Skills/Tech: Python, FastAPI, OpenAI API, LangChain, ChromaDB, Streamlit, Retrieval-Augmented Generation
House Plant Health Classification Using Amazon Rekognition
For this project, I built an end-to-end houseplant health classification pipeline using Amazon Rekognition Custom Labels. Starting with 91 personal houseplant photos, the pipeline achieved 95% accuracy with a precision of 1.00 and recall of 0.80 on the minority unhealthy class.
Key Skills/Tech: Python, Amazon Rekognition, Custom Labels, Amazon S3, AWS Lambda, IAM, Boto3, Albumentations, Scikit-learn
Film Industry EDA (SQL & R)
I built this project to analyze patterns across 60,000+ films using a production-style data pipeline, loading raw TMDB data into a SQLite database, querying it with SQL, and wrangling and visualizing the results in R. Two findings stood out: popularity and rating are essentially uncorrelated (r = 0.087), and Drama movies consistently outrate Action films (6.27 vs. 5.96).
Key Skills/Tech: R, tidyverse, ggplot2, lubridate, ggrepel, SQLite, DBI, RSQLite, Exploratory Data Analysis, Statistical Analysis
Financial Time-Series Forecasting: NVIDIA Case Study (In Progress)
- Objective: Investigate the limitations of classical time-series models under high-volatility market conditions using NVIDIA stock as a case study.
- Key Skills/Tech: Python, ARIMA, ARIMAX, MACD, ADF/KPSS Testing, Ljung-Box Diagnostics, LSTM (PyTorch), RMSE/MAE/R²
- Repo: In Progress
📂 Archived / Past Work
Older projects and exploratory work are available in Archived_Projects for reference.