Machine Learning Engineer

Ayman Mahmud Haque

Machine Learning Engineer

■ *************@*****.*** ■ +1-256-***-**** ■ linkedin.com/in/aymanhaque001/

■ github.com/aymanhaque001 ■ U.S Citizen

SKILLS

Languages & Frameworks: Python, SQL, C++, TensorFlow, Keras, PyTorch, FastAPI, MLflow, Scikit-learn, Langchain. React.js, Docker

Tools & Libraries: Git, Pandas, NumPy, Matplotlib, Seaborn, Power BI, OpenVino, Transformers, Optimum-Intel, VLLM, NLTK, Gensim, OpenCV, Apache Kafka, Linux

Cloud: Azure Data Factory, Databricks, AWS Sagemaker Studio, AWS EC2, Power BI EXPERIENCE

Data Scientist - Tata Consultancy Services @ Intel - Cloud Computing Group (CCG) July 2024 - Present

● Optimization and Benchmarking of Generative AI applications on cutting-edge Intel GPU, CPU, and NPU using OpenVino, Optimum-Intel, PyTorch, ONNX, and TensorFlow.

● Fine-tuning and deployment of models on Intel cloud platform and edge devices (AI PC). Generate reports and visualizations on model performance.

● Working directly with clients to assess and debug issues, ensuring scalability and performance. Data Engineer Tata Consultancy Services @ Northern Trust July 2022 - Dec 2023

● Conduct data cleaning, transformation, and validation to ensure data quality and integrity.

● Develop and maintain robust Data Pipelines using Snowflake, Azure Data Lakes, and Azure Data Factory, optimizing ETL processes for efficiency.

Research Assistant Dr. Jamil Saquer, Computer Science Department, Missouri State University July 2022 - Dec 2023

- Fake Hotel Reviews Detection using Topic Modelling and Convolutional Neural Networks

●Performed Data cleaning and preprocessing on a large corpus. Lemmatizing, POS-tagging, stop words removal, etc.

●Employed Unsupervised machine learning (Latent Dirichlet Allocation) for topic modeling to extract relevant features.

●Generated Word Embeddings with Word2Vec skip-gram from the corpus for training and developed a CNN model.

●Achieved 89% accuracy on 5-fold cross-validation and developed other ML models i.e., TF-IDF + Support Vector Machines, Multinomial Naïve-Bayes, and Gradient Boosting classifier to compare results. EDUCATION

Missouri State University, Springfield, USA Master of Science in Computer Science August 2019 – May 2021 BRAC University, Dhaka, Bangladesh Bachelor of Science in Electrical and Electronics Engineering January 2012- December 2017 PROJECTS

Lung Cancer Detection Pipeline (96% Accuracy) End-to-End ML Deployment TensorFlow, Keras, VGG16, DVC, MLflow, Docker, Flask, GitHub Actions, CI/CD, AWS

● Built a deep learning pipeline to classify lung CT scan images into four classes — squamous cell carcinoma, large cell carcinoma, adenocarcinoma, and Normal — using a public Kaggle dataset.

● Achieved 96% validation accuracy by fine-tuning a pre-trained VGG16 model via transfer learning, customizing the classifier layers for multi-class output with TensorFlow Keras.

● Designed a scalable, production-ready ML pipeline with DVC for data and pipeline versioning and MLflow for experiment tracking and model registry.

● Developed a Flask-based REST API for real-time image classification and containerized the application using Docker for consistent deployment across environments.Deployed the complete solution to AWS EC2, following MLOps best practices for cloud-native, scalable machine learning services.

Identifying Optimal Subgroups of Traumatic Brain Injury (TBI) Patients NumPy, Pandas, Matplotlib, seaborn, weka, sci-kit-learn

●Data Cleaning: encoding categorical features and dropping unnecessary features to improve data integrity.

●K-means clustering was used to cluster the dataset into meaningful groups.

●Visualized the results using TSNE and PCA dimensionality reduction methods. Evaluated clusters with a combination of IVMs, i.e., DB, SI, and CH index

●Employed evolutionary feature selection to reduce dimensionality and increase classification performance. Evaluated accuracy using MLP. Drowsiness Detection for Drivers using Deep learning TensorFlow, Keras, OpenCV, NumPy, Scikit-learn,

●Used OpenCV library to extract images from webcam feed and HAAR cascades to extract regions of interest.

●Performed preprocessing of image dataset to train a deep learning model.

●Developed a convolutional neural network to classify detected images to appropriate classes.

●Designed a system to keep scores and alert the driver when eyes are closed for a particular threshold.

●Evaluated validation accuracy for different CNN architectures, hyperparameter tunings, and environmental and lighting conditions. Stock market trend Predictor using Recurrent Neural Networks

●Performed data preprocessing and cleaning on the Yahoo Financial Time series dataset and arranged it into a recurring dataset for training the LSTM model.

●Developed LSTM network to predict prices for each subsequent day; evaluated for several hyperparameter settings; visualization. Awards

Missouri Outreach Graduate Opportunity (MOGO) Scholarship August 2019 – May 2021 Missouri State University, Springfield, USA

Contact this candidate