Senior Audio ML / Research Engineer

Mowka; Bengaluru; On-site; INR 6,000,000 - 8,000,000; Full time

Job summary

Mowka is seeking a Senior Audio ML / Research Engineer in Bengaluru to own training and fine-tuning for the voice AI platform. You will design and run experiments, develop high-quality training pipelines, and push model quality in a fast-growing startup environment.

You will work with audio, speech, and diffusion-related models, leveraging GPUs and distributed training to scale. 4–5+ years of ML experience and hands-on research are expected.

Qualifications

• Trained or fine-tuned ML models; understands training data, objectives, evaluation.
• Experience with speech/audio/TTS/ASR and related ML problems.
• 4–5+ years of ML engineering or research in fast-paced environments.

Responsibilities

• Train and fine-tune models for real-time voice agents.
• Build and improve training pipelines for large-scale audio datasets.
• Experiment with architectures, objectives, and fine-tuning techniques.
• Implement research ideas and test on real-world problems.
• Analyze failures and improve models via data, training, and experiments.
• Work with GPUs and distributed training to scale model training.

Skills

Model training

Fine-tuning models

Training pipelines

Experimentation

GPUs distribution

Model evaluation

PyTorch

Hugging Face

Tools

PyTorch

Hugging Face

Job description

Senior Audio ML / Research Engineer

• Salary: ₹60L – ₹80L
• Location: Bengaluru Bangalore Urban
• Experience: 5 years
• Job Type: Full Time
• Posted: today

About the job

Hiring: Senior Audio ML / Research Engineer

We're building India's most powerful custom SLMs and real-time Speech-to-Speech voice agents that already handle millions of calls across banking, healthcare, and new economy brands.

We want to take our platform to even crazier scale. We're building things that real people use. Come train and improve models you're proud of.

If you're a Data Scientist / ML Engineer who thrives on pushing the limits of what's possible and wants to own the model training and fine-tuning for an early-stage AI startup, this is your seat at the table.

The Mission

As our Senior Audio ML / Research Engineer, you will own the training and fine-tuning of the models behind our voice AI platform. You'll be responsible for training and improving custom audio and speech models, building high-quality training pipelines, running experiments, and turning research ideas into models that perform reliably in the real world. Your goal is to continuously improve model quality, performance, and the overall product experience.

What You'll Do

• Train and fine-tune models for custom SLMs and real-time Speech-to-Speech voice agents
• Build and improve training pipelines for large-scale speech and audio datasets
• Experiment with model architectures, training objectives, and fine-tuning techniques to continuously improve model quality
• Implement ideas from research papers and evaluate whether they work on our real-world problems
• Analyze model failures and performance and improve them through better data, training, and experimentation
• Work with GPUs and distributed training to efficiently train models at scale

What We're Looking For

Model Training Depth: You've actually trained or fine-tuned ML models. You understand the training process, data, objectives, evaluation, and trade-offs. You're comfortable going beyond simply using or deploying existing Hugging Face models.

Audio / Speech Experience: You've worked with speech, audio, TTS, ASR, Speech-to-Speech, generative audio, or related ML problems. Experience with diffusion-based audio models is a strong plus.

Research & Experimentation: You can take an idea from a research paper, implement it, design experiments, and determine what actually improves the model. You don't necessarily need to invent new architectures, but you should be comfortable going deep into how models work.

Battle-Tested Experience: You have 4-5+ years of serious ML engineering or research experience, ideally in the of early-stage startups or high-velocity product teams.

Owners, Not Employees: You don't wait for a ticket. You identify what needs to improve, run the experiments, and ship models that drive the product forward. You measure success by better models, better performance, and happier users.

The Rewards

• Impact: Your models will directly power millions of voice interactions. You're building the core engine, not a side experiment.
• High-Growth Environment: We're angel-backed and scaling fast. This is a front-row seat to building a category-defining voice AI company.
• Culture: A high-intensity, high-trust team. No politics, no fluff—just building and winning.

Interested?

If you've spent the last few years training models, fine-tuning them, breaking them, figuring out why they failed, and making them better, let's talk.