Fully Funded PhD Position in Secure, Reliable and Efficient Transformer-Based AI Systems
TalTech (Tallinn University of Technology) is inviting applications for a fully funded PhD position in
secure, reliable and efficient transformer-based AI systems
. The project sits in the
Computer Systems Department
and focuses on the intersection of
computer science, computer architecture, hardware security, and AI efficiency
.
The research topic addresses the growing use of transformer-based models such as
large language models (LLMs), vision transformers (ViTs), and vision-language models (VLMs)
. As these models become larger and more resource-intensive, techniques like
quantization, pruning, sparsity, KV cache optimization, compression, and approximate computing
are increasingly important for reducing memory footprint, latency, energy use, and computational cost. This PhD will investigate how such efficiency methods affect the
reliability and hardware security
of transformer inference systems.
The project will examine representative transformer architectures and accelerators, studying resilience under
side-channel attacks, fault attacks, hardware trojans, and soft errors
. A key aim is to understand how optimization techniques change the way corrupted values propagate through models and how auxiliary structures such as scale factors, zero-points, sparsity indices, metadata, and control information can influence failure behavior.
Another major objective is to design
lightweight protection and mitigation mechanisms
that preserve the efficiency advantages of optimized transformer inference. The work will involve balancing trade-offs among
security, reliability, model quality, performance, memory footprint, energy consumption, and hardware overhead
.
Eligibility:
applicants should hold a
Master's degree
in Computer Engineering, Computer Science, Artificial Intelligence, or a closely related field. Strong skills in machine learning/deep learning, Python and PyTorch, computer architecture or digital systems, and English are required. Experience with LLMs, ViTs, quantization, pruning, sparsity, approximate computing, or hardware security/reliability is a plus.
Funding:
the position is fully funded, with salary determined by TalTech regulations and salary tables.
Application:
send a CV and academic transcripts to the listed supervisors by email, using the specified subject line. Shortlisted candidates will be invited to an online interview. The deadline is
16 September 2026
, and the start date is as soon as possible, with some flexibility.
TalTech is Estonia’s leading university of engineering and technology and offers an international research environment in Tallinn, with strong expertise in information technology, cybersecurity, engineering, and digital technologies.
