PinnedAftab Gazali·Jan 20Helios: A Citation-First, Cost-Gated RAG Assistant for Research PDFs (Built with LangGraph)A LangGraph pipeline that routes queries, enforces per-sentence citations, and cuts cost with a two-tier model gate Code + demo…
Aftab Gazali·Jan 22Gilgamesh: Building a Domain QA Assistant on a 4GB GPU (Phi-3 + QLoRA + RAG)Fine-tuning Phi-3-mini with 4-bit NF4 QLoRA + lightweight FAISS RAG for source-grounded Mesopotamian Q&A. Code + demo + results: GitHub…
Aftab Gazali·Jan 21Day-Ahead Load Forecasting in Québec: A Leakage-Safe, Auditable ML PipelineFixed 23:00 cutoff, 24 hourly forecasts, rolling-origin backtests, and a LightGBM “24 specialists” setup that cuts MAE ~66% vs a weekly…
Aftab Gazali·Jan 20From BDD100K to YOLOv8: A Practical Traffic-Safety Detection Pipeline (FiftyOne + Ultralytics)End-to-end detection pipeline on a BDD100K subset for car, person, rider, traffic light — with a baseline run and an improved final model…
Aftab Gazali·Jan 30, 2024Deep Learning for Handwritten Digit Recognition: A PyTorch ApproachHandwritten digit recognition stands as a crucial task within computer vision, finding utility across various domains such as optical…
Aftab Gazali·Jan 25, 2023Conditional GANS(cGANs) for MNIST DatasetGans stands for Generative Adversarial Networks. It is a type of deep learning model composed of two neural networks: a generator and a…
Aftab Gazali·Jan 15, 2023GANs on Hand-Written Digits DatasetGans stands for Generative Adversarial Networks. It is a type of deep learning model that is composed of two neural networks: a generator…
