نبذة عني
MSc Computational Data Science graduate specializing in machine learning, deep learning, and LLM-based applications. Experienced in building end-to-end ML solutions, including model development, evaluation, fine-tuning, …
MSc Computational Data Science graduate specializing in machine learning, deep learning, and LLM-based applications. Experienced in building end-to-end ML solutions, including model development, evaluation, fine-tuning, and deployment using PyTorch, FastAPI, Docker, and Kubernetes. Strong in data preprocessing, feature engineering, performance optimization, and scalable AI systems, with hands-on experience in RAG pipelines, NLP, computer vision, and LLM fine-tuning. Passionate about applying AI to real-world problems and contributing to data-driven engineering teams.
الخبرة
Data Science Intern
<p>Developing a stateless Computerized Adaptive Testing (CAT) engine in R</p><p> Implementing adaptive item-selection and ability-estimation algorithms based on psychometric principles</p><p> Optimizing assessment workflows for efficient and personalized testing experiences</p><p> Integrating the CAT engine with .NET frameworks to enable scalable deployment and seamless integration within educational technology platforms</p>
Data Science Intern
<p>Built and optimized machine learning models using Python and scikit-learn for predictive analytics tasks</p><p> Performed data preprocessing, feature engineering, and exploratory data analysis (EDA) on structured datasets</p><p> Applied hyperparameter tuning and cross-validation techniques to improve model performance and generalization</p><p> Developed visualizations and analytical reports using Matplotlib and Seaborn to communicate insights effectively</p>
Full-Stack Development Intern
<p>Developed and integrated end-to-end application features across frontend and backend systems</p><p> Improved application performance by optimizing API calls and database interactions</p><p> Collaborated on debugging and testing to enhance system reliability and user experience</p>
Data Science Intern
<p>Cleaned and preprocessed datasets using Pandas and NumPy for machine learning tasks</p><p> Engineered features and trained models to improve prediction accuracy</p><p> Evaluated model performance using appropriate metrics and validation techniques</p>
المشاريع
Brain Tumor Detection via Convolutional Autoencoder
<p>Built an unsupervised MRI anomaly-detection model</p><p>using reconstruction-error thresholding and heatmap localization. PyTorch, NumPy, OpenCV, Computer Vision</p>
Malayalam Sentiment Analysis Service
<p>Built and deployed a production-ready sentiment analysis API using FastAPI, Docker, GitHub Actions (CI), and</p><p>Kubernetes. Implemented model serving with health probes, automated testing, containerization, and ingress-</p><p>based routing for scalable ML deployment. FastAPI, Docker, Kubernetes, CI/CD, MLOps, Model Deployment</p>
ChestXray14: 1024 ˜A—1024 VAE + Latent Classifie
<p>Trained a convolutional VAE and used 64-d latent embeddings for multi-label disease prediction with macro F1</p><p>evaluation, reducing storage requirements from 55 GB to 450 MB. PyTorch, scikit-learn, Deep Learning</p>
PDF RAG Assistan
<p>Developed a PDF question-answering system using Retrieval-Augmented Generation with Ollama embeddings</p><p>and ChromaDB for semantic search. Built an interactive Streamlit UI for document upload and contextual query-</p><p>ing. Python, Streamlit, ChromaDB, Ollama, LLM</p>
Personalized Post-Transplant Management AI Agent
<p>Developed a clinician-facing research prototype for kidney-transplant decision support that integrates a Duffingbased</p><p>Dynamic-ML predictive model, uncertainty-aware GFR trajectory simulation, transplant-specific Retrieval-</p><p>Augmented Generation (RAG), a Validator Agent, and a Streamlit dashboard. Built patient-aware literature</p><p>retrieval over PDF/TXT/Markdown sources using ChromaDB and Ollama LLM, with citation validation, deterministic</p><p>safety checks, LLM-based review, and automatic repair for weakly grounded outputs. Validated the predictive</p><p>backbone on held-out N2 cases, achieving ROC AUC 0.840 / F1 0.772 at 365 days and ROC AUC 0.811 / F1</p><p>0.730 at 730 days. Python, Streamlit, NumPy, pandas, ChromaDB, Ollama, RAG, LLM Agents, Nonlinear Dynamical</p><p>Systems, RK4, Clinical AI</p>
Conversational Fine-Tuning of Qwen2
<p>Built an end-to-end supervised fine-tuning pipeline for Qwen2.5-1.5B-Instruct using the Topical-Chat dataset to</p><p>generate context-aware assistant responses from multi-turn dialogue history. Developed custom preprocessing,</p><p>label masking, and training components to resolve instability issues including zero-loss and NaN failures. Imple-</p><p>mented LoRA-based parameter-efficient fine-tuning, checkpointed training, and inference workflows on Kaggle</p><p>GPU in</p>