At Seneca Innovation Center, I design and deploy LLM systems and RAG pipelines, including open-source models on local NVIDIA DGX infrastructure for privacy-preserving inference. I also analyze GPU memory, latency, and model performance to optimize inference.
I built LLM Inspector, an open-source CLI for analyzing GPU memory during LLM inference and estimating potential savings from quantization strategies. It supports vLLM, Hugging Face Transformers, Ollama, and custom FastAPI backends.
Earlier, at Digital Group Performance, I redeveloped an e-tourism website and built an Angular and Spring Boot interface and admin dashboard for booking and billing. I also built EyeMate, a voice-driven assistant that uses a lightweight vision-language model to understand images and read documents aloud.
