A multi-service AI content pipeline that scrapes web pages, extracts content and screenshots, rewrites material with LLMs, supports contextual review, maintains version history through ChromaDB, and uses a reinforcement-learning service for review feedback. The system is containerized with Docker Compose, includes service health checks, automated tests, SSRF protection, CORS origin controls, and persistent vector storage.

A Go-based API gateway designed around a consistent request-processing pipeline covering logging, metrics, tracing, tenant resolution, analytics, rate limiting, chaos injection, and reverse proxying. Redis-backed token-bucket rate limits persist across restarts, while Prometheus, Grafana, and OpenTelemetry provide end-to-end observability. The project includes a live gateway, public Grafana dashboard, metrics endpoints, and controlled latency, failure, and packet-drop experiments.

A published Llama 3.1-based language model artifact focused on legal-domain text generation. The model is distributed in multiple inference-friendly formats and can be served through Transformers, llama.cpp, vLLM, SGLang, Ollama, and Unsloth, including OpenAI-compatible inference endpoints. The repository demonstrates hands-on model packaging and deployment across modern LLM inference stacks.

A hybrid medical-image augmentation pipeline for chronic kidney disease classification using CT kidney imagery. It combines SMOTE for feature-space balancing with an Auxiliary Classifier GAN for synthetic image generation, then benchmarks VGG16, EfficientNetV2, and MobileNetV2 across original and augmented datasets. Reported accuracies range from 95.1% to 99.2%, with training runs documented for reproducibility.

A privacy-first Chrome extension that locally audits installed extensions for potentially risky permissions, broad host access, and known tracker domains. It provides deterministic 0-10 risk scoring, severity indicators, CSV reporting, local persistence through IndexedDB, and direct disable/uninstall actions. The architecture performs analysis entirely on-device with no telemetry or remote calls.