Remote Voice AI Research Engineer - Production-Driven

Voicegain; Polska; Remote; PLN 180,000 - 300,000; Full time

Job summary

Voicegain is hiring a researcher who ships. The role focuses on building speech-to-text and NLP systems that run inside production telephony stacks, handling languages like Spanish, French, Polish, and more.

You will work directly with the founders, formulating problems, reading papers, and deploying robust solutions in real-world settings. Expect deep ML work—transformers, embeddings, and fine-tuning—plus production-ready Python and ability to navigate large Java/C++ codebases.

Qualifications

• PhD in ML/NLP/speech or equivalent research experience.
• Hands-on experience fine-tuning models (LoRA/QLoRA) and rigorous evaluation.
• Deep expertise in ASR, spoken-language understanding, or information extraction.
• Strong Python skills and ability to ship production code in large Java/C++ codebases.
• Experience with production ML infrastructure and NVIDIA stack is a plus.

Responsibilities

• Formulate problems, read papers, analyze deployment constraints, prototype, measure on real data, analyze failures, then build the production version.
• Prototype, evaluate failures, and monitor how it behaves on live traffic.
• Collaborate directly with the founders to ensure a short path from experiments to production implementations.

Skills

ML foundations

LLMs beyond prompting

Speech / NLP depth

Python expert

Java/C++ codebase

ML infra (GPU, Kubernetes)

NVIDIA stack experience

AI tooling fluency

Independent decision-maker

Education

PhD in ML/NLP/Speech

Tools

NeMo

Riva

Triton

TensorRT

LoRA/QLoRA

Kubernetes

Job description

Voicegain is hiring a researcher who ships. The role focuses on building speech-to-text and NLP systems that run inside production telephony stacks, handling languages like Spanish, French, Polish, and more.

You will work directly with the founders, formulating problems, reading papers, and deploying robust solutions in real-world settings. Expect deep ML work—transformers, embeddings, and fine-tuning—plus production-ready Python and ability to navigate large Java/C++ codebases.