I build complete, deployed AI applications — from model training through to production APIs and web interfaces. My work spans computer vision, NLP, and classical ML, with a focus on shipping working systems, not just notebooks.
- Computer Vision: Deepfake detection, image classification, and neural style transfer using EfficientNet, ResNet50, and VGG19.
- NLP: Aspect-based sentiment analysis using transformer models (DistilBERT, zero-shot BART).
- Classical ML: Diagnostic and predictive models using Random Forest and XGBoost, with an emphasis on interpretable, real-time scoring.
- Full-Stack Deployment: FastAPI backends, Next.js/React frontends, deployed on Vercel and Render.
- Deepfake Detector — Classifies uploaded face images and video as real or AI-generated, with a confidence score. Trained an EfficientNet-B0 model on 100k+ images; full-stack app with a FastAPI backend and Next.js frontend, deployed on Render + Vercel.
- NeuralLens — A ResNet50-powered image classification app supporting real-time recognition across 1,000+ object categories.
- SentimentSense — An NLP dashboard performing aspect-based sentiment analysis using DistilBERT and zero-shot classification.
- Diabetes Detector — A diagnostic tool using a Random Forest Classifier to predict diabetes risk from clinical metrics.
- AetherQuant — A crypto market trend classifier using XGBoost, analyzing technical indicators via live market data.
- Portfolio: Portfolio Hub
- LinkedIn: linkedin.com/in/syed-m-ali-faraz
- Upwork: Freelance Profile
"A good developer knows the math behind the code."