🎓 Discover PhD and Master's programmes at leading universities worldwide — Sign up free to save searches and get email alerts
UNI

Explainable AI Guided Hardware Acceleration for Energy-Efficient AI Systems

University of Maine Electrical and Computer Engineering
✓ Fully Funded 🎓 Artificial Intelligence 🎓 Computer Science 🎓 Electrical Engineering energy efficiency reinforcement learning large language models explainable ai hardware acceleration approximate computing sustainable computing

Explore designing energy-efficient hardware to boost explainable AI models like large language models. Engage in cutting-edge research integrating AI, hardware design, and sustainability. Join a collaborative lab with strong mentorship at University of Maine.

AI-generated overview

🌍
Why This Research Matters

This research enhances the deployment of trustworthy AI by enabling systems that are both powerful and transparent. Energy-efficient hardware reduces environmental impact while meeting increasing computational demands, critical for adoption in healthcare, finance, and autonomous technology sectors.

SecureAISNN Approximate computing EDA

Project Description

Project Overview

This PhD research is situated in the GENIUS Laboratory at the University of Maine, targeting Explainable AI (XAI) Guided Hardware Acceleration. The goal is to develop novel energy-efficient hardware solutions that facilitate transparent and interpretable AI by leveraging approximate computing techniques. This cutting-edge project supports AI models including large language models (LLMs) and reinforcement learning.

What You Will Do

The successful candidate will work closely with Prof. Ayesha Siddique and the GENIUS Lab team to design and prototype hardware accelerators that improve the computational efficiency of explainable AI. Research will include advanced programming, AI model analysis, and hardware-software co-design, focusing on sustainability and reliability.

Expected Outcomes

You will contribute to advancing academic knowledge in explainable AI while producing tangible hardware designs that promote greener AI technologies. Deliverables include publications, tools for AI interpretability, and practical hardware demonstrations suited for real-world sensitive applications.

Why This Matters

Explainable AI is critical for trustworthy deployment of AI in safety-sensitive domains. Hardware acceleration addresses computational challenges by reducing energy usage and latency, thus enabling scalable, reliable AI. This research furthers sustainable AI innovation with potential impacts in healthcare, finance, and autonomous systems.

Entry Requirements

Strong background in computer science or computer engineering with excellent results in B.Sc. or M.Sc. degrees. Essential knowledge in artificial intelligence and machine learning. Highly valued experience in large language models and reinforcement learning. Proficiency in C, C++, Python, graph theory, dynamic programming, and optimization.

How to Apply

Refer to the LinkedIn post by Prof. Ayesha Siddique for application details: https://www.linkedin.com/posts/ayeshasiddique-as_deepneuralnetworks-llm-explainableai-share-7453558380553568256-d403

Eligibility

UK/Home
EU
International

Supervisor Profile

PA
Prof. Ayesha Siddique
University of Maine, Electrical and Computer Engineering

Prof. Ayesha Siddique leads research on Explainable AI Guided Hardware Acceleration, focusing on developing sustainable and interpretable AI systems using approximate computing. Her work bridges artificial intelligence, hardware design, and efficient computing technologies. She is an emerging expert in AI transparency and hardware-based acceleration techniques.

Key Publications

2021 21 citations
Exploring fault-energy trade-offs in approximate dnn hardware accelerators
2023 20 citations
Exposing reliability degradation and mitigation in approximate DNNs under permanent faults
2022 18 citations
Is approximation universally defensive against adversarial attacks in deep neural networks?
2018 14 citations
Approxcs: Near-sensor approximate compressed sensing for iot-healthcare systems
2023 11 citations
Improving reliability of spiking neural networks through fault aware threshold voltage optimization

Research Contributions

Studied fault-energy trade-offs in approximate DNN hardware accelerators to optimize reliability and energy use.
Helps improve the design of energy-efficient and reliable AI hardware.
Investigated reliability degradation and mitigation techniques in approximate DNNs under permanent faults.
Contributes to making approximate neural networks more robust and dependable.
Analyzed the defense effectiveness of approximation against adversarial attacks in deep neural networks.
Informs the development of secure and resilient AI models against adversarial threats.
Proposed near-sensor approximate compressed sensing methods for IoT healthcare systems.
Enables more efficient and low-power data acquisition and processing in healthcare IoT.

Related Opportunities

Undergraduate/Graduate Research Assistantship on AI and Machine Learning for Protein Modeling
Auburn University at Montgomery Dr. Sutanu Bhattacharya 🎓 Artificial Intelligence 🎓 Computational Biology

Explore AI applications in protein modeling and bioinformatics. Develop machine learning solutions with Python and contribute to advancing biomedical research. Gain valuable experience working under an NSF Expand AI gra…

This research improves protein structure predictions, crucial for drug discovery and understanding diseases such as COVID-19. Enhancing AI …

300+ citations · h12
Computational biology Bioinformatics Machine Learning Data Science
Rigorous Safety and Reliability in Autonomous Systems via Formal Verification and Data-Driven Control
University of Birmingham Prof. Sadegh Soudjani 🎓 Applied Mathematics 🎓 Computer Science Deadline: 10 May 2024

Explore how to develop mathematically rigorous methods ensuring safety and reliability in autonomous systems by integrating control theory, formal verification, and probabilistic approaches. Ideal for candidates eager t…

This research is crucial for advancing the safety and reliability of autonomous systems deployed in real-world safety-critical applications…

3500+ citations · h30
Cyber-Physical Systems Safe Autonomy & AI Model Checking Formal Methods
Building Secure and Trustworthy Autonomous Systems
University of Arkansas Dr. Hasan Shahriar 🎓 Artificial Intelligence 🎓 Computer Science

Explore security and trust issues in autonomous systems including robots and vehicles. Develop resilient cyber-physical systems using advanced AI and privacy techniques. Address real-world challenges in infrastructure p…

Research in secure and trustworthy autonomous systems is critical as these technologies become integral to transport, robotics, and infrast…

Additive Manufacturing Optimization modeling
PhD in Computer Architecture and High-Performance Digital Circuit Design for Edge AI Computing
INRS University Dr. Shervin Vakili 🎓 Computer Engineering 🎓 Electrical Engineering Deadline: 15 May 2026

Explore novel designs in digital and computer architecture to boost edge AI computing. Develop hardware-aware machine learning techniques to optimize circuits for performance and efficiency in embedded systems.

This research addresses critical needs for energy-efficient and high-performance AI hardware at the edge, enabling real-time processing in …

300+ citations · h14
Computer Architecture High-Performance Architectures for Real-time Embedded Systems Hardware