Ashhadul Islam

Ashhadul Islam

Harnessing AI to bridge visionary innovation with real-world impact.

Bridging Worlds: Navigating the Frontiers of Academia and Industry

My journey has been a dynamic interplay between the rigor of academia and the practical demands of industry. With a foundation in advanced AI research, I have consistently sought to apply cutting-edge methodologies to real-world challenges, whether it be optimizing complex systems for global enterprises or pioneering new technologies in emerging startups. This narrative weaves together the experiences, innovations, and insights gained from navigating the evolving landscapes of both academia and industry.

Academic Background

  • Master’s Degree in Engineering - IIIT Bangalore
    Completed my Master’s degree with a focus on [specific area of study if applicable].

  • PhD in Artificial Intelligence and Machine Learning - Hamad Bin Khalifa University, Doha, Qatar
    Moved to Doha to pursue a PhD in Artificial Intelligence and Machine Learning. My PhD work focused on two main areas:

    • Optimizing Deep Neural Networks: During the pre-LLM era, I worked on reducing the size of large models like VGGNet without compromising accuracy. This work was published in ICML workshops and other significant conferences.
    • Handling Imbalanced Data: Developed a variant of SMOTE called KNNOR to address bias in classifiers by generating new data. I created a Python library for this technique, which can be easily installed and used like other scikit-learn libraries. This work was published in the Applied Soft Computing journal in 2022, and the paper has since garnered 60 citations.

Industry Experience

  • AXA Insurance Company, Belgium (Worked from India)
    Worked as an IBM Mainframe Operator, monitoring and handling the abnormal executions of jobs.

  • Data Scientist - Development Bank of Singapore
    Transitioned to the role of Data Scientist where I focused on ATM Demand forecasting using SARIMA and ARIMA models.

  • Intern - National Instruments
    Developed Python wrappers for C# code during my internship, gaining valuable experience in integrating different programming languages.

  • Data Engineer - Redhat
    Briefly worked as a Data Engineer, a role I secured through competitions and hackathons, just like my previous position at the Development Bank of Singapore.

  • Consultancy Role - Malaysia
    Worked on a short-term consultancy project for a company in Malaysia, applying LLMs to create on-premises document retrieval systems. Despite limitations like internet restrictions and resource constraints, we successfully established a prototype for document search within the company’s documentation.

Startups And COnsultancy Experience

  • Hipla.io, Singapore
    Worked on building a Computer Vision & Surveillance management system. I implemented various models for tasks such as social distancing monitoring, occupancy monitoring, and hard-hat monitoring in construction sites. The project was developed using Flask and React, and deployed as Docker containers on clients’ on-premise servers to process up to 20 CCTV feeds simultaneously.

  • Health Startup, Australia
    Assisted in processing health sensor data from fitbits. I built a system on AWS Kinesis to stream data into AWS DynamoDB, applying different ML models. I created a custom pipeline on an AWS EC2 server for retraining models on daily data, rather than using AWS Sagemaker.

Post-PhD Work

  • Postdoctoral Researcher - KTH Royal Institute Of Technology, Stockholm, Sweden
    Guiding and advising research teams, including PhD scholars and Master’s students, and collaborating with industry partners. My work includes:
    • Symbolic regression for modeling lung pressure and ventilator dynamics.
    • Combining physiological modeling and machine learning to develop patient-specific optimization strategies for mechanical ventilation.
    • Enhance precision medicine in critical care.
  • Postdoctoral Researcher - Hamad Bin Khalifa University, Doha, Qatar
    Guiding and advising research teams, including PhD scholars and Master’s students, and collaborating with industry partners. My work includes:
    • Evolving my algorithm for oversampling data from classification to regression, which was published in Expert Systems with Applications this year.
    • Applying neural network pruning and optimization methods to small LLMs, aiming to reduce model size without losing accuracy.
    • Building an intra-university RAG-based LLM chatbot with a knowledge base of around 32,000 articles and 250 ebooks, now actively used by university researchers.

Future Aspirations

I am now seeking to harmonize the strengths of industry, research, and academia to develop scalable solutions that not only address complex challenges but also contribute to a more sustainable and accessible world.