┌─────────────────────────────────────────────────────────────────┐
│ visitor@github:~$ whoami │
│ > Muneeb Ahmad — AI & ML Engineer │
│ │
│ visitor@github:~$ cat profile.json │
│ > edge AI · agentic systems · computer vision · data science │
│ > 3+ yrs exp · final year CS · fiverr level 1 seller │
│ > building things that actually run in the real world │
└─────────────────────────────────────────────────────────────────┘
|
Spatial assistance system for blind and low-vision users. On-device inference only — no cloud, no latency. Dual-stream depth + semantics fused with quantized Gemma 4.
|
AI alignment framework for surgical knowledge removal from LLMs. High-precision LoRA-masking erases targeted data or behaviors without touching base capabilities.
|
||||||||||||||||
|
RAG pipeline for financial document intelligence — intelligent chunking, vector retrieval, and LLM reasoning for high-accuracy financial Q&A. |
Distributed recommendation engine using MFCC audio feature extraction and Apache Spark's ALS collaborative filtering with MongoDB as the metadata layer. |
||||||||||||||||
|
Async OCR desktop app for images + PDFs via Tesseract, wrapped in a PyQt6 interface with theming, file history, and drag & drop support. |
Real-time attention monitor via CV — tracks attention, fatigue, frustration, and engagement as live percentages with personalized AI coaching messages. |
||||||||||||||||