ML/AI Engineer (ASR) Offline

🟠 Details: we are looking for an ML/AI Engineer (ASR) to fine-tune and optimize an Automatic Speech Recognition system for Swiss German. In this role, you will work hands-on with PyTorch and Hugging Face to adapt pretrained ASR models, build reproducible training pipelines, and improve recognition accuracy through data quality refinement and controlled experimentation (WER/CER).

🟠 About the project: our client is advancing a speech recognition solution with a strong focus on dialect adaptation and real-world audio robustness. The current phase centers on dataset preparation and validation, model fine-tuning (e.g., Whisper), decoding and language-model optimization, and data-centric quality improvements to achieve production-grade transcription accuracy across diverse speakers and conditions.

🟠 Qualifications:

• Solid experience in automatic speech recognition (ASR) systems.
• Practical experience with model fine-tuning and evaluation, including metrics such as WER and CER.
• Strong hands-on proficiency with PyTorch and the Hugging Face ecosystem.
• Good understanding of decoding strategies (e.g., beam search) and integration of language models.
• Familiarity with data-centric AI approaches and dataset quality improvement practices.
• English language proficiency at B2 level (upper-intermediate) and above.

🟠 Would be a plus:

• Experience adapting models for multilingual or dialectal speech.

🟠 Responsibilities:

• Prepare and validate speech datasets, including audio—text alignment, normalization, and dataset splitting.
• Build and maintain reproducible training pipelines using PyTorch and Hugging Face tools.
• Fine-tune pretrained ASR models (e.g., Whisper) for target data and use cases.
• Apply established ASR training techniques such as SpecAugment, curriculum learning, and decoding optimization.
• Improve pseudo-label filtering workflows and overall training data quality.
• Design and run controlled experiments to evaluate models and reduce WER.
• Deliver reproducible training configurations, model checkpoints, and clear technical documentation.

🟠 Interview steps:

• Screening with Recruiter (about 30 minutes).
• Interview with Project Manager (about 30 minutes).
• Technical interview with ML Engineer (about 30 minutes).

🟠 By joining us, you will get:

• 15 paid vacation days + 30 unpaid days off per year.
• 4 paid public holidays (New Year, Independence Day, Day of the Defender of Ukraine, and Christmas).
• Accounting and legal support (corporate accountant & lawyer).
• Flexible hours, remote work, and quality equipment.
• 100% certification coverage (on success).
• 50% coverage for additional training.
• Internal workshops and knowledge sharing.
• $50/month wellbeing budget (50% co-pay) for sport, education, therapy, and massage.
• Team-buildings, gifts, and company events.
• Internal currency program for extra perks and rewards.

🟠 Our office:

• FuturaHUB, located in Lviv (Kulparkivska 200a).
• An inspiring, modern workplace with a great atmosphere and development culture.
• Conference rooms.
• “Happy Hours” parties.
• Yoga and sports training in FuturaHUB.
• Chill-out area: gaming zone with PlayStations, table tennis & table soccer.
• Beer, coffee, tea, cookies, etc.

🟠 Some of our Values:

• No overtime and pressure.
• Regular rest and burnout control.
• Clean relationships within the company and with the client.
• Responsibility for one's own development, ideas, and interests, as well as for the problems that stand in the way of this development.
• Openness to the new, readiness for changes.