PhyML Invos
Hi, I'm Hasham Akram, a Data Scientist and ML Engineer.
.png)
Hi, I'm an AI/ML Engineer driving impactful solutions with expertise in NLP, Computer Vision, and Generative AI,
leveraging MLOps and data-driven Machine Learning. I focus on orchestrating cutting-edge AI systems dynamically.
Professional Journey:
• Currently working as AI & ML Engineer at Sprouto Groups, leading document classification and automation initiatives
• Previously served as Machine Learning Engineer at AXON Technologies, developing ML models on AWS with significant performance improvements
• Gained experience through internships at CodSoft and iNeuron.ai, developing fraud detection and energy efficiency models
Key Achievements:
• Built AI-Human document classification pipeline achieving 90% accuracy with fine-tuned Gemma-3B and BERTa models
• Led 100% automation of social media lead-generation, reducing manual intervention by 90%
• Developed ML models improving R² by 30% and reducing complexity by 40%
• Cut fraudulent transactions by 30% (saving $1M/year) and lowered customer churn by 20% (saving $500k/year)
My passion lies in applying AI to solve complex problems, combining my physics background with cutting-edge ML techniques
to unlock new insights and drive innovation. I enjoy sharing my knowledge through technical blogs and collaborative projects.
Leading AI/ML initiatives with focus on document classification, automation, and scalable ML solutions.
Developed and deployed ML models on AWS, focusing on disease prediction and AI agent development.
Developed multiple ML models for fraud detection, spam classification, and customer churn prediction.
Focused on energy efficiency prediction and data-driven decision making for construction and structural planning.
an English-to-Urdu translation application built using a Sequence-to-Sequence (Seq2Seq) model with Gated Recurrent Units (GRU) and Bahdanau Attention mechanism.
This project aims to develop a Next Word Prediction Model using stacked Bidirectiona LSTMs and GRU to enhance text input efficiency and user experience.
This project implements a Region-based Convolutional Neural Network (RCNN) model for detecting wheat crops in images. The model is designed to assist in agricultural monitoring, enabling farmers and researchers to assess crop health and density efficiently.
Developed a machine learning model to detect fake news using various machine learning algorithm, achieving an accuracy of more than 92%. The system is trained on a dataset of news articles that have been preprocessed and count vectorized.
This project implements a web application for classifying chicken diseases using deep learning image recognition. The application leverages Keras' pre-trained VGG-16 model built upon TensorFlow to achieve high accuracy in disease detection.
Developed a machine learning model to predict Energy Efficiency for an Indusrial company, achieving an accuracy of 92%. The project involved data cleaning, feature engineering, model training, and deployment using Flask and Docker.
An introductory guide to PINNs, a regularization technique incorporating ODs/PDEs as physical laws for better Convergance toward optimal solution.
So, what is a physics-informed neural network? A great Source of SciML is here.
Description of the certification.
Description of the certification.
Explored astronomical technologies and scientific methods.
Developed and Deployed ML models:
Churn Prediction Model
SMS/Email Spam Classification Model
Fraudulent Transaction Prediction Model
End-to-End Structural Energy Efficiency prediction Modelling and Evaluation
Learned data analysis techniques for astronomical research.