AI Developer - Signal Processing and Speech Technology, Limassol, Cyprus

We are recruiting, on behalf of our client, for an experienced AI Developer  -  Signal Processing & Speech Technology , for their team in  Limassol, Cyprus . About the company: Our client is a subsidiary of a powerful network of high-tech companies and one of the leading international providers of high-quality communications intelligence systems. With nearly 70 years of industry experience, the group provides customers with innovative systems and products that help make the world a safer place. About the role: You'll develop AI systems that go beyond Jupyter Notebooks. The focus is on selecting, training, and optimising models for analysing RF signals and speech data, and then deploying them as high-performance, production-ready applications. Latency, throughput, and resource consumption aren't footnotes here, they're core requirements.  Important note: Only applications from EU nationals who hold a valid working permit in Cyprus can be considered. While a hybrid working scheme for eligible positions is available, all employees must be based in Cyprus. New joiners are required to work onsite from the Limassol office for the first 3–6 months for training and team-building purposes. The work centres on two technical areas: RF signal analysis: Classification, detection, and characterisation of high-frequency signals in broadband data streams, often near real time. Speech processing: Recognition (speaker recognition / diarisation), denoising, and speech enhancement, as well as automatic speech recognition (ASR) under real-world, noisy recording conditions. You'll also help design and develop components of a real-time platform for the automated interception, analysis, and visualisation of radio signals. Responsibilities: Select, adapt, and fine-tune models for signal and speech processing (including CNN/Transformer architectures, state-space models, and end-to-end ASR) Optimise inference: quantisation, pruning, kernel fusion, batching strategies, mixed precision, TorchScript/torch.compile, and export to ONNX/TensorRT Implement high-performance inference paths in PyTorch and LibTorch (PyTorch C++) Deploy and operate models in the NVIDIA Triton Inference Server, including model repository, ensembles, business logic scripting, dynamic batching, and custom backends Build processing chains around the models: signal pre-processing, feature extraction, streaming pipelines, and pre-/post-processing Take prototypes from development to market readiness: benchmarking, profiling, test coverage, reproducibility, monitoring, and model versioning Collaborate closely with system architecture and product management Evaluate new methods and tools for their practical suitability against performance goals Requirements: BSc / MSc in Computer Science, Electrical Engineering, Physics, Mathematics, or a comparable qualification, or equivalent practical experience Several years of experience with PyTorch beyond training: you know where…