Hi,

I am Muhammad Asad Majeed

Software Engineer

London, United Kingdom

Profile Image

About me

I’m an aspiring Software Engineer with a strong interest in AI systems, and research. I enjoy building full-stack applications that merge clean design with intelligent, data-driven functionality, combining backend logic with user-oriented thinking

Currently, I’m pursuing a BSc in Computer Science at University College London (UCL), where I’m deepening my understanding of software engineering principles, algorithms, and data structures. My coursework has equipped me with a solid foundation in computer science theory and practical programming skills.

My technical experience includes working with Python, Java, C, and various web technologies. Recently, I’ve been planning to work on a research project focused on benchmarking large language model agents. I plan on exploring how different agents perform across structured tasks, and building evaluation tools to assess their reasoning, consistency, and adaptability. I also plan to build a master agent that evaluates user requirements and chooses the best agent for the job. This project will deepen my interest in AI alignment and agentic behavior, while also strengthening my understanding of evaluation frameworks and reproducible research.

Outside academics, youl will either find me on a football field or glued to my laptop trying to build things from scratch, working on open-ended projects, and learning through hands-on exploration. I’m particularly excited about opportunities that combine software engineering with machine learning, and I’m always open to collaborations in research or product-focused applications of AI.

Skills

Python
JavaScript
Java
HTML5
CSS3
React
Node.js
Git
Docker
SQL
Machine Learning
AI

Projects

Project 5

ForeSight

An AI-powered geopolitical risk platform built to find the "whispers" that precede major market shifts. Foresight uses the Perplexity API to analyze thousands of obscure global sources, connecting events like local strikes to their multi-million dollar impact on specific company supply chains, automating an analyst's work in seconds.

Node React Tailwind CSS Perplexity
Project 4

Finance Tracker

A full-stack AI-powered Finance WebApp using React, Node.js, PostgreSQL, and AWS to securely connect bank accounts, monitor spending, and display real-time data. Integrated Plaid API and JWT authentication, delivering a complete cloud-based software solution from frontend to infrastructure.

Node React CSS3 PostgreSQL PostgreSQL
Project 3

AI Job Application Assistant

An AI-powered co-pilot that automates personalized job applications using an advanced Retrieval-Augmented Generation (RAG) pipeline. It intelligently analyzes job descriptions to strategically select and showcase your most relevant skills and project experiences.

Python HTML5 FastAPI FASTAPI CSS3 JavaScript
Project 1

WonderRoute

An AI-powered multi-user trip planner that generates personalized, photo-spot-aware itineraries by syncing locations, preferences, and ratings using Google Maps and Gemini AI.

React Node.js Google APIs
Project 2

Personal Portfolio

A clean, responsive personal portfolio showcasing my projects, skills, and experiences as a Computer Science student. Built to reflect both my technical abilities and design sense using modern web technologies.

HTML5 CSS JavaScript

Github

GitHub Stats
GitHub Streak
Most Used Languages

GitHub Contributions

GitHub Contribution Chart

Publications

Benchmarking AI-Agent frameworks

arXiv preprint (forthcoming), 2025

Majeed M.A, Maaz M.

A research paper benchmarking leading AI agent frameworks like LangChain, AutoGen, and CrewAI, with a focus on tool integration, retrieval strategies, and multi-agent collaboration. The study evaluates performance across real-world tasks, highlighting trade-offs in scalability, latency, and context handling.

Read Paper

Benchmarking Modern LLM Architectures: Transformers and Beyond

arXiv preprint (forthcoming), 2025

Majeed M.A., Maaz M.

A research paper benchmarking diverse LLM architectures including Transformers, Mixture of Experts, State Space Models (e.g. Mamba), and Recurrent Memory Models (e.g. RWKV). The study compares their performance across NLP tasks, focusing on accuracy, latency, scalability, and context retention.

Read Paper

Recommendations

Working with Asad on our machine learning project was incredibly productive. He brings both technical expertise and creative thinking to every challenge.

MM

Muhammad Maaz

AI/ML Researcher, University of Augsburg

Contact

Send me a message