Master Thesis – Designing Ternary Neural Networks for Edge Vision (AI)

Location: Lund, Sweden; on-site

Company

hirify.global develops network technology and security solutions including video surveillance, access control, intercom, audio systems, and intelligent analytics applications.

What you will do

• Design and train ternary neural networks with weights constrained to −1, 0, and +1 for edge-vision tasks.
• Assess ternary weights in existing object-detection models and adapt network architectures and training methods to retain useful accuracy.
• Investigate network blocks, quantization-aware training, activation precision, and mixed-precision designs.
• Benchmark ternary models against suitable higher-precision and INT8 baselines.
• Evaluate accuracy, model size, CPU latency, throughput, and, where equipment is available, energy use on real edge hardware.
• Consider the potential relevance of compact low-precision models to future compute-in-memory hardware; working with such hardware is not required.

Requirements

• Master’s student affiliated with a Swedish University/College, either directly or through an exchange program.
• Studying Computer Science, Electrical Engineering, Computer Engineering, or a related field.
• Strong interest and curiosity in deep learning, computer vision, model compression, embedded systems, or computer architecture.
• Programming skills, preferably in Python.
• Experience with PyTorch or a similar deep-learning framework.
• For proposals involving two students, apply in pairs and identify the co-applicant in the application.
• Submit a CV, cover letter, and University/College grade summary.

Culture & Benefits

• Collaborative work within a network technology company with approximately 6,000 employees in more than 50 countries.
• Work on security and intelligent analytics solutions for a smarter and safer world.
• Applications are accepted in Swedish and English.
• The position is a fixed-term thesis worker role.
• Background checks may apply; candidates will be notified before any checks are conducted.

Hiring process

• Apply through the thesis proposal advert.
• Submit one application per student when applying as a pair and clearly identify the co-applicant.