Projects
Telegram SentinelPython, Node.js, FAISS, LLM
- Architected a multi-service Telegram intelligence platform combining Python-based message ingestion with a Node.js API backend.
- Implemented FAISS-powered vector search for fast and accurate contextual retrieval from large message corpora.
- Integrated LLM-driven response generation to transform raw Telegram data into human-readable insights within seconds.
Bash-AIShell-Scripting, Ollama, CLI commands
- Designed an AI-powered Linux assistant using Ollama for real-time command recommendations and execution.
- Created dynamic model selection (0.5B / 1.5B / 3B) based on available system RAM to optimize performance.
- Integrated secure and interactive CLI with command confirmation and error handling to ensure safe operations.
Contract Summarizing ChatbotNext.js, FastAPI, LLM
- Devised a chatbot capable of summarizing 500+ pages of contracts easily within 3 minutes.
- Enabled information retrieval using FAISS with 90% accuracy, processing each document in seconds.
- Combined FAISS and Llama 3 to generate accurate responses from vast legal documents in under 5 seconds.
Cloud Storage WebsiteLoad Balancing, Reverse Proxy
- Implemented load balancing with weighted round-robin for efficient traffic distribution.
- Developed reverse proxy with route-specific load distribution and rate limiting.
- Enhanced security and performance with authentication middleware and rate limiting.