AI-ML Engineer – VoiceArena

Location: Gurgaon (On-site)

Type: Full-time

Experience: 1–2 Years

About VoiceArena

At VoiceArena, we believe voice will become the primary interface between humans and machines. We build benchmarks, datasets, and AI systems that power some of the world's most widely used speech models.

Our mission:

- Enable machines to communicate as naturally as humans.
- Build benchmarks and datasets that form the backbone of global speech AI progress.
- Drive improvements through high-quality, diverse, real-world data — not just compute.

What You'll Work On

This is not a conventional AI/ML engineering role. You will work on real-world speech AI challenges, build scalable systems, and directly contribute to work that influences the global speech AI ecosystem.

Data Preparation & ML Infrastructure

- Design, build, and maintain end-to-end data pipelines, from raw audio ingestion to structured, training-ready datasets.
- Develop reliable, versioned, and reproducible workflows for large-scale speech data processing.
- Optimize data pipelines for efficiency, scalability, and quality.
- Collaborate with research and engineering teams to ensure seamless integration between data infrastructure and model training workflows.

Benchmarking & Evaluation

- Design, develop, and maintain evaluation frameworks for Automatic Speech Recognition (ASR) and speech-to-speech systems.
- Benchmark leading speech models to identify real-world strengths, weaknesses, and failure modes.
- Build automated evaluation pipelines to measure model performance across languages, accents, noise conditions, and speaking styles.
- Analyze evaluation results and generate actionable insights to improve speech AI systems.

Modeling & Fine-Tuning

- Fine-tune and optimize speech recognition models such as Whisper, wav2vec2, HuBERT,



and other open-source architectures.
- Work towards achieving low Word Error Rates (WER), targeting approximately 5% on suitable datasets and evaluation conditions.
- Experiment with multilingual, code-switched, accented, and noisy speech data.
- Develop and improve training, validation, and inference pipelines for speech models.
- Work with large-scale, production-grade speech datasets spanning 20+ Indian languages

Engineering & Deployment

- Write clean, productive, and production-ready Python code for ML workflows.
- Build and optimize model inference pipelines for performance and scalability.
- Use cloud infrastructure and containerization tools to deploy and manage ML workloads.
- Troubleshoot technical challenges across data processing, model training, and evaluation systems.
- Take ownership of engineering tasks from experimentation through implementation.

Who We're Looking For

Required Qualifications

- 1–2 years of hands-on experience in AI/ML engineering, preferably in speech AI, NLP, audio processing, or related areas.
- B.Tech/B.E. in Computer Science, Electrical Engineering, AI/ML, or a related technical field.
- Strong proficiency in Python and experience with PyTorch, TensorFlow, or Hugging Face Transformers.
- Practical experience building, training, fine-tuning, or evaluating machine learning models.
- Familiarity with speech recognition architectures such as Whisper, wav2vec2, HuBERT, or similar models.




- Experience working with large datasets, data preprocessing, and ML pipelines.
- Understanding of machine learning fundamentals, model evaluation metrics, and experimentation workflows.
- Working knowledge of Git, Linux, and cloud platforms such as AWS, GCP, or Azure.
- Strong problem-solving skills and the ability to independently execute technical projects.

Preferred Qualifications

- Hands-on experience with speech recognition, speech-to-speech systems, or multilingual AI models.
- Experience fine-tuning models on Indian languages, code-switched speech, or low-resource datasets.
- Familiarity with ASR evaluation metrics such as Word Error Rate (WER) and Character Error Rate (CER).
- Experience with Docker, Kubernetes, distributed training, or GPU-based ML infrastructure.
- Contributions to open-source AI/ML projects, published research, or strong GitHub portfolios.
- Experience optimizing model inference, training efficiency, or data processing pipelines.

Why Join VoiceArena?

- Work on meaningful AI problems: Contribute to speech AI systems, benchmarks, and datasets used to evaluate and improve leading AI models.
- Build at scale: Work with diverse speech datasets across 20+ Indian languages and challenging real-world conditions.
- Own your work: Take responsibility for building, experimenting, evaluating, and improving ML systems.
- Learn through execution: Work closely with engineers and researchers solving complex speech AI challenges.
- Make a real impact: Your work will directly contribute to advancing speech AI for millions of users.

If you're an engineer who enjoys building ML systems, experimenting with speech models, and solving challenging real-world problems, we'd love to hear from you.

-

📌 AI-ML Engineer (Gurugram)
🏢 Voice Arena
📍 Gurugram