Meghna Choudhury

Results-driven data engineering professional with solid foundation in designing and maintaining scalable data systems. Expertise in developing efficient ETL processes and ensuring data accuracy, contributing to impactful business insights. Known for strong collaborative skills and ability to adapt to dynamic project requirements, delivering reliable and timely solutions.

• Music: Piano, Bass guitar
• Drawing and Painting: Charcoal art, digital painting, watercolor painting
• Founded and led the IT & Technology Club in high school, promoting innovation and peer learning through events and workshops
• Represented peers as part of the Directorate of Student Affairs, helping deliver the cultural and technical festival
• Organized events and cultural initiatives as part of the Fun Committee in my first work organization

1. FoodVantage: Multimodal AI Grocery Health App for diet awareness -2026
Built a Streamlit-based grocery health app using Gemini vision/agent workflows, DuckDB, Plotly, and Open Food Facts for real-time item scanning, metabolic scoring, allergy detection, nutrition retrieval, personalized insights, and AI meal planning.

2. International Student AI Consultant for Education and Career Pathways (In Progress) -2026
Designing a FastAPI-based multi-agent RAG system for program matching, visa guidance, and career analysis for international students, with retrieval over official datasets, vector indexing of visa/regulatory PDFs, evaluation/logging, and a dashboard prototype.

3. Data Analysis and Visualization of Cosmic Deep Space -2025
Conducted EDA and data wrangling in R using tidyverse and ggplot2 on the JWST UNCOVER DR3 Abell 2744 photometric catalog (~74K sources), quantifying missingness and completeness across NIRCam filters and building hexbin and dot-density visualizations to analyze brightness-redshift trends and spatial non-uniformity.

4. FXCurrencies Model -2025
Developed a Python-based USD/CHF FX trading framework on ~6 years of Bloomberg OHLC data, combining optimized exponential-smoothing crossover strategies (grid-searched α/β, confirmation/gap filters, momentum exits) with ML-based probabilistic signal models including KNN, Random Forest, XGBoost, and neural networks; evaluated performance using time-series splits, trading simulations, and risk-return metrics such as Sharpe, CAGR, and max drawdown.

5. Machine Learning Clustering for Colorectal Cancer Tissue Patches -2022
Performed unsupervised model selection on 5,000 colorectal cancer tissue patches across 9 tissue types, benchmarking K-Means, GMM, Hierarchical Clustering, and Louvain over multiple feature spaces; compared deep embeddings from ResNet50, InceptionV3, VGG16, and PathologyGAN with PCA/UMAP reduction and evaluated cluster quality using Silhouette Score and V-measure.

6. Baxter Robot Playing Chess -2022
Simulated a Baxter robot to detect, pick, and move chess pieces, modeling human-like spatial reasoning and move execution with 90% accuracy.

7. Navigation Hand Device for the Visually Impaired -2021
Built a real-time LiDAR-based navigation aid in embedded C++, integrating servo control and STL-based spatial mapping for obstacle-aware guidance.

8. Humanoid Kid-Size Virtual Soccer Simulations -2021
Developed a virtual soccer team in ROS using behavior-based decision algorithms for attacking, defending, and goalkeeping, achieving 90% accurate gameplay in simulation.