ML Engineer Audio Real-time Production

#vacancy #English #hot #job #Kazakhstan #Uzbekistan #Georgia #Belarus #ML #Audio Elinext is an IT consulting company that have been delivering software development services on time and budget since 1997. We are searching a ML Engineer to join our team. As an engineer at Elinext, you'll collaborate with different teams like product, analytics, and operations on code that empower us to iterate quickly, while focusing on delighting our customers. You'll be working on our innovative project focused on real-time audio source separation and multi-sound recognition. Project: Real-time system capable of simultaneously identifying and separating multiple sound sources (vocals, instruments, etc.) from live music streams with low latency. This role is strongly focused on production-grade, real-time audio ML, not offline experimentation. Requirements: Core Audio ML: Proven experience with real-time audio source separation (Demucs, Hybrid Demucs, Band-Split RNN, TF-GridNet); Strong understanding of multi-label audio classification — simultaneous detection of multiple sound sources in a mixed signal; Deep knowledge of audio feature extraction: mel-spectrograms, STFT, CQT, MFCCs, chromagrams; Experience with streaming/chunked audio processing with low-latency constraints. Deep Learning & Architectures: Proficiency in PyTorch (preferred) or TensorFlow for audio tasks; Hands-on experience with architectures proven in audio: U-Net, Wave-U-Net, Conv-TasNet, Audio Spectrogram Transformer (AST), HTS-AT; Understanding of attention mechanisms and transformer-based approaches for audio; Experience with model quantization, pruning, and ONNX/TensorRT export for real-time inference. Real-Time Processing: Experience building low-latency audio pipelines (target latency < 200ms); Knowledge of streaming inference: overlapping windows, buffered processing, causal convolutions; Familiarity with WebSocket / gRPC streaming for audio data; Understanding of trade-offs between latency, accuracy, and computational cost. Audio Engineering Fundamentals: Proficiency with librosa, torchaudio, soundfile, scipy.signal; Understanding of sample rates, windowing, hop lengths, and their impact on real-time performance; Experience handling various audio formats and codec considerations. Infrastructure & Deployment: Experience deploying audio ML models to production (GPU inference servers, edge devices, or cloud); Familiarity with NVIDIA Triton, TorchServe, or custom serving solutions; Proficiency with Docker, CI/CD for ML pipelines; Monitoring and logging for real-time ML systems. English: English language at Intermediate and above level is a must. Benefits: broad responsibility, autonomy and visibility in an engineering role; in-depth exposure to real-world customer issues across a global customer base; small-company feel in a growth environment; extensive, invaluable exposure, and experience to top-notched, leading-edge technologies in Cloud computing, home monitoring systems, and a vast of other exciting, hot products; working in a friendly environment with a team of creative and enthusiastic engineers; ability to promote and try out cutting-edge technologies for the app development; retirement plan contributions matching (applicable to country of residence); health benefits (applicable to country of residence).