I am a Machine Learning Engineer with a strong interest in designing, training, and deploying end-to-end ML systems. My work focuses on transforming raw data into reliable models and deployable services, with particular emphasis on Deep Learning, Computer Vision, and Natural Language Processing (NLP).
I place strong value on understanding models beyond high-level APIs, including implementing core architectures from scratch to gain deeper insight into their design, efficiency trade-offs, and limitations. I am especially interested in LLMs, model efficiency, and MLOps-driven workflows.
- 🔭 Currently working on Agentic AI systems and refining production-ready deep learning pipelines
- 🌱 Deepening expertise in Large Language Models (LLMs), Transformer internals, and prompt engineering
- 👯 Open to collaboration on applied ML, MLOps, and LLM system design
- 💬 Ask me about Machine Learning, Deep Learning, Computer Vision, NLP, MLOps, TensorFlow, and PyTorch
- 📫 How to reach me: abdulrasheedolakiitan@gmail.com
Programming Languages
Machine Learning & Deep Learning
Data Science & Analysis
MLOps & Deployment
Databases
Version Control
Built a complete computer vision pipeline for maize leaf disease classification under real-world field conditions.
- Worked with a large-scale dataset of 18,148 real-world images
- Implemented LFMNet, a lightweight multi-attention CNN optimized for background noise and subtle inter-class differences
- Achieved 86% accuracy in 5 epochs, emphasizing architectural efficiency
- Used DVC for data and experiment versioning and Docker for reproducible deployment
- Exposed inference through a Flask-based API
Tech: Python, TensorFlow, scikit-learn, DVC, Docker, Flask
🔗 https://github.com/AbdulRasheed6/end-to-end_mazie_disease_classification
Developed an end-to-end retrieval-augmented generation (RAG) medical chatbot for querying medical documents.
- Semantic retrieval using
all-MiniLM-L6-v2embeddings - Vector search powered by Pinecone
- Prompt-based response generation using Gemini Pro
- Dockerized deployment on AWS EC2
- CI/CD pipeline using GitHub Actions, Amazon ECR, and secure secret management
Tech: LangChain, HuggingFace, Pinecone, Flask, Docker, AWS
🔗 https://github.com/AbdulRasheed6/end-to-end_Medical_chatbot
This project analyzes Saudi Aramco’s stock performance under major macroeconomic and geopolitical disruptions, including the COVID-19 pandemic, the Russia–Ukraine conflict, and global oil price movements (OPEC basket).
The goal was to understand how external socio-political shocks affect short- and medium-term market behavior, supporting better-informed planning and risk mitigation rather than speculative long-term forecasting.
- Analyzed historical Saudi Aramco stock data alongside oil price indicators
- Studied the relationship between geopolitical events, market volatility, and stock movement
- Focused on short-term predictive signals and trend sensitivity
The workflow followed a structured ETL (Extract, Transform, Load) pipeline:
- Extract: Stock prices, oil prices, and market indicators
- Transform: Data cleaning, feature engineering, and temporal alignment
- Load: Prepared datasets for modeling and evaluation
Tech: Python, Pandas, scikit-learn, TensorFlow, Matplotlib
🔗 https://github.com/AbdulRasheed6/Aramco_stock_analysis-Forecasting.git
Implemented core Transformer-based language models from scratch to develop a deep understanding of architecture, training dynamics, and efficiency trade-offs.
- Implemented tokenization, causal self-attention, multi-head attention, and autoregressive generation
- Focused on attention masking, weight tying, and training loop design
Tech: Python, Pytorch 🔗 https://github.com/AbdulRasheed6/GPT2.git
- Implemented RMSNorm, Rotary Positional Embeddings (RoPE), and Grouped-Query Attention (GQA)
- Explored efficiency-driven architectural choices for large-scale autoregressive models
- Emphasized correctness and architectural clarity over API abstraction
Tech: PyTorch, Python
🔗 https://github.com/AbdulRasheed6/Custom_Llama2.git