Research Profile
Shahmir Khan
CS Graduate | AI & ML Researcher | LLM Applications
Computer Science graduate with distinction, focused on deep learning, recommender systems, and LLM-powered applications that solve real-world problems.
Research Profile
CS Graduate | AI & ML Researcher | LLM Applications
Computer Science graduate with distinction, focused on deep learning, recommender systems, and LLM-powered applications that solve real-world problems.
Hi, my name is Shahmir Khan. I have completed my Bachelors in Computer Science with a distinction. During my bachelor's I focused on core subjects including programming fundamentals, data structures and algorithms, artificial intelligence, and parallel & distributed systems — which helped me understand how scalable systems are designed and implemented, with a strong focus on LLM-powered AI applications that solve real-world problems.
My academic focus spans the intersection of deep learning, intelligent systems, and human-centered AI.
Explored Neural Collaborative Filtering (NCF) combining MLP and GMF to overcome limitations of traditional CF approaches — cold-start, sparsity, and linear MF assumptions. Results demonstrated strong non-linear pattern capture on implicit feedback data.
Read Thesis →Hands-on experience fine-tuning LoRA adapters on AMD MI300X GPUs using ROCm. Built conversational AI agents leveraging Gemma-4 via Fireworks API for wellness and recommendation use cases.
Academic grounding in designing scalable distributed architectures. Interested in how distributed compute enables large-scale model training and inference pipelines.
Applied classical ML (KNN, Random Forest) to heart disease prediction with EDA-driven insights. Interested in extending to deep learning models for clinical decision support.
Final Year Project
Neural recommendation engine fusing MLP and GMF (Neural Collaborative Filtering) for an e-commerce platform. Achieved 89% accuracy — a 10% gain over traditional baseline models. Handles cold-start and data sparsity by learning from implicit feedback.
Hackathon
Conversational wellness AI built at a hackathon. Fine-tuned a LoRA adapter on AMD MI300X GPUs (ROCm) using a custom 39-example dataset. Integrated cycle intelligence and emotional wellness agents using Gemma-4 via Fireworks API.
AI Agent
An intelligent loan processing agent that automates and streamlines loan application workflows using AI-powered decision making and document analysis.
Trained KNN and Random Forest classifiers for heart disease prediction. Evaluated using accuracy, AUC/ROC, confusion matrix, and classification report. Performed EDA uncovering sex vs. heart disease correlations in the dataset.
Competitions where ideas meet execution under pressure.
Built a wellness AI agent leveraging AMD MI300X GPUs on the AMD Developer Cloud with ROCm. Fine-tuned a LoRA-adapted LLM on a custom wellness dataset and integrated multi-agent reasoning for personalized health recommendations.
Stay tuned for upcoming participation and results.
Recognition and milestones along the way.
Graduated with distinction, demonstrating strong academic performance across core CS subjects.
To be added — this section will be updated with certifications, competition wins, and recognitions.
To be added soon.
A summary of my education, skills, and experience.
Once available, you can preview or download the full CV directly from this section.
Download CVFeel free to reach out for research collaborations, opportunities, or just a chat.