Hi, I'm Abdur Raafay
Computer Systems Engineer & AI Researcher specializing in Edge AI, local LLMs, and autonomous multi-agent systems.
About
Powered by the synergy of code, embedded hardware, and intelligence. I am a Computer Systems Engineering graduate from NED University of Engineering and Technology specializing in Edge AI and agentic workflows.
My research and development focus on building optimized, fully localized AI pipelines. Whether quantizing LLMs to run offline on low-power devices like the Raspberry Pi 5 or engineering complex self-correcting agent teams using LangGraph, I love pushing the boundaries of what local hardware can achieve.
What I can do
Edge AI & Embedded Systems
Architecting fully offline, private AI applications on constrained hardware. Expertise in model quantization, hardware optimization on Raspberry Pi 5, and localized vector search.
Multi-Agent Systems & RAG
Designing autonomous multi-agent pipelines with LangGraph and LangChain. Implementing automated code execution, feedback loops, and semantic codebase indexing.
Computer Vision & Detection
Training and benchmarking deep learning models (RT-DETR, YOLOv10, Faster R-CNN) for complex object detection, evaluating accuracy-latency trade-offs.
Robotics & embodied AI
Developing control scripts and algorithms for physical systems. Interfacing sensor suites with edge processors (Raspberry Pi/microcontrollers) and deploying basic embodied AI behaviors.
Work Experience
AI Intern
Pakistan International Airlines
Built an AI-powered automated monitoring system to track employee compliance and workstation usage across multiple screens, optimizing security and operational efficiency.
Data Science Intern
Neurocomputation Lab, NED University
Developed data processing tools and machine learning pipelines to clean, analyze, and model complex research datasets, accelerating lab research workflows.
Check out my latest work
I've worked on a variety of projects, from low-power wearable edge devices to complex multi-agent coding engines. Here are a few of my favorites.
Neuraid: Personal Cognitive Memory Assistant
A fully localized, offline wearable cognitive assistant utilizing optimized LLMs, local speech-to-text/text-to-speech, and Chroma vector databases on a Raspberry Pi 5 to ensure absolute user privacy.
Agentic Technical Research Paper Analyzer
A full-stack agentic research assistant that parses PDFs, indexes pages with ChromaDB, and coordinates a multi-agent pipeline (summary, methodology, critic) to generate citation-grounded reports.
Autonomous EDA Agent
An autonomous multi-agent data analytics platform engineered with LangGraph. Converts natural language queries into executable data engineering scripts with automated self-correction loops.
YouTube Q&A Chatbot
A conversational YouTube assistant built with Chainlit. Fetches transcripts, indexes video segments in a local FAISS vectorstore, and uses LangChain to answer natural language questions with source grounding.
Grocery Object Detection & Benchmarking
Fine-tuned and benchmarked state-of-the-art vision models (RT-DETR, YOLOv10x, and Faster R-CNN) on a custom multi-class grocery dataset, evaluating accuracy-latency profiles.
Deep Codebase Intelligence Assistant
An intelligent repository parser mapping multi-file codebases into hierarchical vector spaces. Offers context-aware vector retrieval workflows for semantic search and tracing.
Tech Publications
I write detailed guides sharing my research, FPGA design implementations, and end-to-end hardware-software engineering workflows.
Building a 4-Bit Adder (Gate Level) on the Nexys A7
A comprehensive guide to building a 4-bit ripple-carry adder from the ground up using gate-level Verilog. Walks through defining submodules, writing XDC constraint files, and running synthesis and implementation in Xilinx Vivado to run on physical hardware.
Interactive FPGA Design: Driving the Nexys A7 RGB LED
Implementing a real-time RGB LED color controller on the Artix-7 FPGA using Verilog and Vivado. Covers reading push-button input vectors, handling state declarations to control tri-color diodes, and mapping port signals to hardware pins with XDC constraints.
Education & Certifications
Bachelor of Engineering
Computer Systems Engineering
NED University of Engineering and Technology
Oct 2022 – Jun 2026
Karachi, Pakistan
Certifications & Courses
-
Certified Agentic and Robotic AI Engineer Program
PIAIC — Ongoing
-
CS50’s Artificial Intelligence with Python
Harvard University
-
Certified Data Scientist
NED Academy
Technical Skills
Programming
Frameworks & Libraries
Core Areas
Tools & Hardware
Contact
Feel free to reach out. I am open to discussing technical collaborations, research opportunities, or software engineering positions.