Senior Speech & Voice Biometrics Engineer

Quick Facts

• Role: Senior Speech & Voice Biometrics Engineer

• Focus: Voice biometrics for contact-center fraud detection; R&D through production

Description

You will lead the scientific and algorithmic development of a new voice biometrics product—from alpha versions to production deployment—for fraud detection in contact centers. The work includes designing, training, tuning, and optimizing speaker verification/identification models, adapting models to real telephone audio, and building scoring and calibration methods with evaluation using biometric metrics (FAR/FRR/EER, DET). You will also develop real-time verification and implement anti-spoofing and voice deepfake detection, while mentoring the team.

Responsibilities

• Drive end-to-end R&D: from alpha releases to production deployment in voice biometrics

• Train and optimize speaker verification/identification models (e.g., ReDimNet, ECAPA-TDNN, ResNet x-vectors)

• Adapt open-source models to telephone audio conditions (noise, codecs, channel effects, devices)

• Design scoring and calibration methods and select thresholds per customer

• Define and run evaluations using FAR, FRR, EER and DET; present results to technical and non-technical audiences

• Develop real-time verification decisions during conversations (IVR short utterances, VAD, diarization)

• Research and implement anti-spoofing and voice deepfake detection

• Evolve an existing production system together with engineering and business teams

• Mentor the team via pair programming, code review, documentation, and knowledge transfer

• Deliver services fully remotely

Requirements

• Several years of commercial experience in speaker verification/identification and voice biometrics projects

• Practical speech and signal processing expertise, including telephone audio and codecs

• Experience training and fine-tuning speaker embedding models (e.g., ReDimNet, ECAPA-TDNN, ResNet x-vectors)

• Deep knowledge of scoring and calibration (AS-norm, S-norm, cohort selection, CSLS, LLR) and thresholding

• Strong understanding of biometric metrics and testing under very low false-accept rates (FAR, FRR, EER, DET)

• Experience with real-time systems plus VAD, diarization, and speaker change detection knowledge

• Very good knowledge of Python and PyTorch

• Familiarity with SpeechBrain, WeSpeaker, NVIDIA NeMo, or Kaldi

• Experience working on existing production systems

• Mentoring experience (pairing, reviews, documentation, knowledge sharing)

• Ability to explain complex topics to non-technical people

• English C1+ (speaking and writing)

Benefits

• Fully remote work

• B2B rate up to 170 PLN/hour + VAT