AARJAV DESAI

MSc Artificial Intelligence student at Queen Mary University of London with hands-on experience in machine learning, NLP, and data-driven problem solving. Previously worked in software quality engineering, where I used automation and analytical thinking to improve system performance and reliability. Now looking for AI/ML internship opportunities to apply my technical skills to real-world projects and continue growing in applied machine learning.

• Natural Language Processing Specialisation - Coursera
• Programming in Python - Coursera
• Data Structures in C/C++ - Udemy
• Effective Problem Solving and Decision Making - Coursera

Business Booster Award - Won the business booster award at Infra Market for contributing to a successful tool using NLP and automation techniques to effectively cut costs and save manpower.

• Sentiment Analysis on Twitter Data (SVM Classifier)

Built an SVM-based sentiment classifier using custom feature extraction (BoW, n-grams, weighted features). Improved preprocessing and optimised model performance through hyperparameter tuning and feature selection. SVM hyperparameters and performing feature selection, achieving a significant performance improvement over the baseline.

• Search & Pathfinding Algorithms

Implemented BFS, DFS, UCS, and A* to solve routing problems on graph networks. Designed heuristics for optimal search, analysed performance, and ensured cycle-free path constraints across all algorithms

• Text Network Analysis on 1984

Constructed a directed word graph from Orwell’s 1984 to compute longest paths, longest quotes, and high-cost sequences. Built heuristic search methods to generate coherent sentence completions and starters.

• Sushi Go! AI Agent (TAG Framework)

Developed a search-based game-playing agent using MCTS/RMHC/heuristic methods. Tuned decision-making strategies, evaluated performance across 100+ simulated games, and conducted ablation studies on algorithmic improvements.

• Distributional Semantics (Word Similarity)

Implemented distributional semantic models to analyse word similarity using context-based features, embedding vectors, and similarity metrics.