AI-driven code modernization studio with 14 specialist agents running in parallel across a DAG orchestration. Converts legacy codebases into modern, runnable target stacks.

• Agent Specialization: Each agent focuses on a specific modernization concern (security, performance, architecture, testing, etc.)
• Real-time Streaming: Server-Sent Events (SSE) pipeline for live agent reasoning
• Multi-LLM Fallback: Automatic cascade from Claude → Gemini → Groq → Ollama with zero downtime
• RAG Engine: ChromaDB stores AST-generated code summaries with SHA-256 fingerprinting to avoid hallucinations
• Export Formats: DOCX reports, Markdown guides, and JSON archives

Unified Angular + FastAPI system for AI-powered media enhancement and narrative generation.

• Facial Enhancement: GFPGAN-powered face restoration from low-quality surveillance footage
• Vehicle Deblurring: DeblurGANv2 for license plate and vehicle detail restoration
• Video Narratives: Automated incident report generation from security camera footage
• VoiceBot UI: Interactive voice-controlled interface for media queries
• Git Submodules: Clean dependency management for rapid local prototyping

Flask-based system that transforms surveillance video into structured law enforcement reports.

• Frame Extraction: OpenCV for frame-by-frame analysis
• Visual Captioning: ViT-GPT2 vision-language model for frame descriptions
• Audio Transcription: Whisper for speech-to-text from video audio
• Report Generation: Groq LLaMA 3 synthesizes general and formal law-enforcement narratives
• Structured Output: JSON with timestamps, transcript, and dual narrative formats

RAG-powered HR automation that extracts skills from resumes into structured matrices.

• Privacy-First: Fully local LLM (Mistral via Ollama) — zero candidate data sent to external APIs
• RAG Pipeline: ChromaDB reduces hallucination to near-zero vs baseline LLM-only approaches
• Document Parsing: Supports PDF and DOCX resumes with robust text extraction
• Skill Mapping: Semantic matching for skill categorization and normalization
• Session Management: UUID-based session storage with rotating log handlers

Fully autonomous AI agent for generating, producing, and publishing educational videos.

• Script Generation: Gemini generates educational scripts on demand
• Audio Synthesis: gTTS converts scripts to natural-sounding voiceovers
• Video Composition: MoviePy combines frames, overlays, and audio
• Auto-Publishing: YouTube API integration for direct uploads with metadata
• GitHub Actions: Runs entirely on serverless cron — no infrastructure needed
• Custom Thumbnails: Dynamic thumbnail generation with TextClip overlays

Live interactive career profile showcasing skills, projects, timeline, and achievements as an engaging single-page experience.

• Canvas-Based: Performance-optimized interactive visualization
• Responsive Design: Works across desktop and mobile browsers
• Career Timeline: Visual representation of professional journey
• Project Showcase: Clickable project cards with links and descriptions
• Skill Matrix: Interactive visualization of technical expertise areas

Research integration for practical face restoration algorithms.

• Real-world Applications: Surveillance footage enhancement, legacy photo restoration
• Deep Learning: State-of-the-art neural network architectures for blind face restoration
• Integration: Core technology used in AriesGPT for facial enhancement

Computer vision utility for extracting, processing, and analyzing individual video frames.

• Batch Processing: Efficient frame extraction for large video files
• Format Support: Works with common video formats (MP4, AVI, MOV, etc.)
• Integration Point: Foundation layer for video analysis projects like AriesGPT and Video to Narrative

Complete ecosystem for AI-assisted application and workflow modernization.

• Backend: FastAPI for high-performance async operations
• Frontend: TypeScript/React for responsive UI
• AI Integration: Multi-agent system for modernization analysis
• Enterprise Ready: Production-grade error handling and logging