Private-First RAG system bridging static local documentation with Mistral-7B/Llama-3 LLMs. Uses FAISS-powered vector DB for zero-leak, secure local data querying.
Dual-engine NDA risk analyzer — 150+ regex patterns across 12 legal categories + Google Flan-T5 AI summarization. Red-team tested against adversarial contract variants.
High-impact developer tool that programmatically identifies, summarizes, and injects docstrings into your Python code using CodeGPT/CodeT5 AI models and AST parsing.
Multimodal AI story generator synthesizing rich prose instantly. Decouples local computations using Qwen-72B & Gemini-3.0 cloud fallbacks for low-compute real-time scripts and Edge Neural vocal acting.
Advanced image-to-image virtual staging app using Hugging Face's serverless backend and Qwen-Image-Edit-2511. Automatically refurnishes empty spaces with smart scaling and client-side SSR fallbacks.
Vintage roleplay museum summoning historical figures from FAISS vector stores. Enforces era-specific boundaries via Llama-3, memory windows, and a custom Streamlit parchment UI.
Multi-modal Discord bot — text chat, image generation, audio transcription & vision AI. Powered by Ollama with Mistral, Stable Diffusion & Faster-Whisper running 100% locally.
Autonomous pipeline: RSS ingestion → LLM curation via Ollama → voice digest with TTS → formatted HTML email briefings. 5 categories, zero human input.
AI-powered code reviewer using Google Gemini 2.5 Flash via LangGraph. Quality scores, severity-based issues, fix suggestions & session email reports. CI/CD deployed.
Seamless fabric texture generator using Neural Circular Padding to eliminate seams, combined with Real-ESRGAN for 4K manufacturing-ready upscaling.
Highly-optimized RAG engine indexing PDFs, DOCX, and TXT files instantly into a FAISS Vector Space. Streams cognitive critiques using Llama 3.3 (Groq API) to map skill gaps.