Unclaimed profileBuilt from public GitHub and Hugging Face activity. Awais-icarus has not joined Lounge yet.Claim it
Not on Lounge yet
Awais-icarus
developerAI/ML Intern @ NVIDIA · Building local LLMs & AI dev tools · 2x Hackathon Winner · CSE '28, Presidency University
Public projects
AMcoli
AMcoli is a C++ inference engine designed to run massive Mixture-of-Experts (MoE) LLMs on consumer hardware. It streams expert weights from NVMe SSDs on-demand, caching active layers in VRAM and RAM to bypass system memory limits. The runtime features auto-hardware detection, built-in model downloading, and a prefetching cache simulation.
C++
ARIA-IDE
🌌 ARIA: A premium, AI-native desktop IDE built with Electron & React. Features multi-model side-by-side comparisons, an agentic pipeline (Planner/Coder/Reviewer), local/cloud LLM integrations, and a sleek, modern pure-black editor canvas.
HTML
ResolveIQ
ResolveIQ is an autonomous AI customer support ecosystem for B2B SaaS. Built on Google Cloud, it uses Gemini, Gemma, and LangGraph to auto-resolve tickets via Vertex AI RAG. It features a self-learning loop, Agent-to-Agent (A2A) incident detection, AI Studio deduplication, BigQuery analytics, and an MCP server for extensibility.
Python
SARA-Distillation
Locally-hosted AI assistant built via knowledge distillation — a 500M-param Qwen2.5 student distilled from 4B+ ▎ teacher models, served through a Flask app with voice I/O (Whisper STT / Kokoro TTS), live talk mode, and HTML ▎ artifact rendering
Python
ML-LAB
Weekly Learning
Jupyter Notebook
Likely stack
LangGraphWhisperMCPReact
