3 hours ago
Singapore, SingaporeSenior
Responsibilities
- Develop streaming multilingual speech recognition and keyword-spotting systems for far-field single-microphone camera audio.
- Build voice activity detection, end-of-turn detection, turn-taking, barge-in, and voice-interaction systems.
- Develop noise suppression, echo cancellation, dereverberation, and target-speaker-extraction models.
- Develop sound event detection, audio classification, and audio captioning systems.
- Run end-to-end audio experiments covering data preparation, training, fine-tuning, ablations, and evaluation.
- Build evaluation harnesses using WER, CER, miss and false-alarm rates, F1, mAP, end-of-turn latency, PESQ, DNSMOS, and echo-cancellation metrics.
- Build audio data pipelines with licensed corpora, synthetic mixing, room impulse responses, camera noise, augmentation, and labeling workflows.
- Design, compress, and benchmark small streaming models using pruning, distillation, and int8 quantization for camera chips.
- Work with firmware engineers to integrate models into camera products and report reproducible experimental results.
- Reproduce speech and audio AI research, test published baselines on company data, and communicate findings to technical and non-technical stakeholders.
Requirements
- Master's degree or PhD in Computer Science, Electrical Engineering, Signal Processing, Acoustics, Artificial Intelligence, or a related field focused on speech or audio.
- Hands-on experience training and evaluating deep-learning models for audio signal-level or audio-understanding tasks.
- Strong knowledge of audio signal processing, including STFT, filterbanks, mel features, pitch, harmonics, room acoustics, reverberation, SNR, and level.
- Familiarity with audio front-end processing, including noise suppression, echo cancellation, dereverberation, gain control, and voice activity detection.
- Ability to design and train small streaming models within compute, memory, and latency budgets and evaluate model size, MACs, and real-time factor.
- Familiarity with speech recognition, sound event detection, audio classification, audio captioning, pretrained audio encoders, audio-language models, and LLM integration.
- Experience with multilingual speech recognition, WER and CER scoring, text normalization, voice interaction, end-of-turn detection, and turn-taking is highly valued.
- Experience with audio augmentation, synthetic mixing, room impulse responses, speed perturbation, SpecAugment, speaker/device/room-based test splits, and task-specific metrics.
- Strong programming skills in Python and PyTorch, with familiarity with Linux, multi-GPU training, experiment tracking, and Git.
- Experience with far-field audio, on-device deployment, quantization-aware training, ONNX, TFLite, sherpa-onnx, vendor NPU toolchains, voice agents, or audiovisual models is a plus.
Benefits
- Medical benefits are provided.
- Yearly bonus and performance bonus are provided.
- The role is based in an office near Tai Seng MRT in Singapore.
- The posting specifies work days per week but does not state the number of days.
Categories
About Reolink
Reolink designs and sells security cameras and systems for homes and small businesses, including Wi‑Fi, PoE, 4G and dual‑lens models, NVRs, and smart video doorbells, with mobile apps and optional cloud storage. Founded in 2009 and headquartered in Wilmington, Delaware, it is privately held. The company distributes globally and serves use cases from indoor and outdoor monitoring to off‑grid, solar‑powered deployments.