AI research. Software delivery. Applied impact.

Benchmark vision-language and generative AI models on HiPerGator GPU clusters to evaluate automated hallucination detection.

Previous role: Machine Learning Researcher
Trustworthy Engineered-Autonomy Lab ? Sept 2025 ? Apr 2026

• Designed and executed experiments generating videos with Wan 2.1 T2V 1.3B and HunyuanVideo from T2VCompBench and ViBe benchmark prompts, systematically quantifying semantic discrepancies between prompts and generated content.
• Developed a fine-grained hallucination taxonomy identifying object omissions, attribute mismatches, spatial relationship errors, and semantic drift across video generation stages.
• Benchmarked automated hallucination detection using Qwen3-VL and other VLMs on severity-level classification; evaluated using Balanced Accuracy, Macro F1, AUROC, and AUPRC.

• Built an end-to-end pipeline that captures video/image streams, reconstructs them into 3D assets using 2D Gaussian Splatting (2DGS), and integrates them into Unity for real-time interactive visualization.
• Designed and trained neural reconstruction models (3DGS/2DGS, COLMAP/GLOMAP) on multi-view data to generate high-fidelity meshes and point clouds; selected 2DGS for its superior exportable mesh quality over 3DGS.
• Accelerated large-scale reconstruction and rendering using NVIDIA B200 GPU clusters with CUDA nightly builds, significantly improving training throughput and pipeline scalability.
• Investigated real-time vs. offline processing tradeoffs to optimize computational efficiency, reducing reconstruction latency and enabling practical deployment workflows.

• Conducted statistical analysis and forecasting on time series forecasting models such as Autoregressive Integrated Moving Average (ARIMA), Vector Autoregression (VAR), and Vector Error Correction Model (VECM) and non-time series models, such as Lasso and Ridge regression, to determine optimal female incumbency.
• Developed ETL pipelines in Python to predict salary costs using workforce datasets containing 1,000+ employee records, improving model accuracy by 20% through exploratory data analysis, feature engineering, and data quality improvements.
• Applied statistical techniques to support HR strategy and workforce planning initiatives.

• Built and maintained responsive web interfaces in Next.js, integrating REST APIs documented with Swagger to enable seamless frontend-backend communication over PostgreSQL databases.
• Developed reusable, scalable UI components with robust state management patterns, reducing code duplication and improving long-term maintainability of the application.
• Contributed to the React Native mobile app, implementing new features and participating in frontend architecture discussions to ensure consistency across web and mobile platforms.