Job Type
Full-time
Description
Propio is hiring a
Senior Machine Learning Engineer, Speech & LLM Training Data to transform large volumes of multilingual conversational audio into high-quality training and evaluation datasets. This hands-on role owns audio processing, dataset curation, annotation and QA workflows, model training, and evaluation for our multilingual speech, translation, and conversational AI systems.
Key Responsibilities:- Define the data roadmap for multilingual speech, translation, multimodal LLMs, and conversational AI.
- Build audio-processing pipelines covering resampling, channel handling, VAD, diarization, language identification, transcription, alignment, and quality filtering.
- Build dataset pipelines for cleaning, deduplication, PII/PHI redaction, quality scoring, sampling, balancing, versioning, and lineage.
- Design annotation guidelines, QA rubrics, golden datasets, and reviewer workflows.
- Build evaluation datasets, analyze model failures, and translate performance gaps into targeted data improvements.
- Run training, fine-tuning, post-training, and evaluation experiments, including SFT, preference data, DPO/RLHF-style workflows, and synthetic data generation.
- Productionize secure, traceable, and reproducible data and ML workflows on AWS.
Requirements
Qualifications:- Bachelor's or Master's degree in Computer Science, Machine Learning, Data Science, Electrical Engineering, Computational Linguistics, or a related field, or equivalent practical experience.
- 5+ years of experience in ML engineering, speech/audio ML, ML data engineering, NLP, or LLM training-data workflows.
- Strong hands-on experience with Python, SQL, Linux, Git, and Docker.
- Experience training or evaluating models using PyTorch, Hugging Face, or comparable ML frameworks.
- Experience with FFmpeg and audio-processing libraries such as TorchCodec, torchaudio, librosa, or equivalent tools.
- Experience with speech-processing tasks such as VAD, diarization, ASR, forced alignment, language identification, and audio-quality analysis.
- Experience with Databricks/Spark, Parquet/Arrow, and large-scale dataset pipelines.
- Working knowledge of AWS S3, SageMaker, Glue, Step Functions, IAM, and KMS.
- Experience with an annotation platform such as Labelbox, Label Studio, Scale AI, Prodigy, Argilla, or custom internal tooling.
- Experience with experiment tracking and data versioning tools such as MLflow, Weights & Biases, DVC, Delta Lake, or LakeFS.
- Experience with multilingual speech, translation, annotation workflows, and evaluation datasets.
Preferred Qualifications:- Experience with multilingual telephony, healthcare, interpretation, or call-center audio.
- Experience with tools such as Silero VAD, pyannote, WhisperX, NeMo, Kaldi, or equivalent speech technologies.
- Experience with distributed processing or training using Ray, PySpark, or similar frameworks.
- Experience with HIPAA, PHI/PII redaction, and secure data governance.
- Experience with low-resource languages, accents, dialects, and code-switching.
- Experience with synthetic data, active learning, weak supervision, or LLM-as-judge evaluation.
#LI-JS1
Notice of AI Use in Job Application ReviewAs part of our commitment in creating a fair, efficient, and consistent hiring process we may use artificial intelligence (AI) to help our recruiting teams organize, summarize, and analyze information provided by candidates, including resumes, application responses, and other materials submitted during the application process.AI may be used to identify patterns, highlight relevant skills, and experience, and assist in comparing a candidate's qualifications with the requirement of a specific role. These tools are to improve efficiency and consistency while supporting more informed hiring decisions, which will ultimately be made by the hiring team.
