TERRA-CD: A Benchmark Dataset for Semantic Change Detection
Published:
Created a benchmark dataset of 5,221 paired Sentinel-2 images and trained Siamese networks for change detection.
Published:
Created a benchmark dataset of 5,221 paired Sentinel-2 images and trained Siamese networks for change detection.
Published:
Built a lightweight Deep Learning framework from scratch using NumPy, featuring a dynamic computational graph engine and reverse-mode differentiation.
Published:
Engineered a graded word-sense plausibility pipeline using DeBERTa, RoBERTa, and GPT-4 to quantify semantic ambiguity.
Published:
Benchmarked Mamba (Selective SSM) and Transformer architectures to investigate trade-offs in computational complexity and memory scaling.
Published:
Engineered a full-stack RAG agent on Cloudflare Workers that acts as a personalized mock interviewer. Features an automated Python evaluation pipeline achieving 90% retrieval accuracy.
Published:
Engineered a serverless AI agent on Cloudflare Workers that converts natural language to SQL and performs real-time sales forecasting.
Published:
Developed a real-time generative prototyping tool mirroring Adobe Firefly, optimizing Stable Diffusion to <1.2s latency using Latent Consistency Models.