Sarvam
Sarvam Machine Learning Engineer
Sarvam AI is developing foundational AI models specifically optimized for Indian languages and cultural use cases.
Get represented
5 minutes to evaluate. 6 months of representation.
What is Sarvam AI?
Sarvam is building a 'sovereign' full-stack AI for India, featuring 'Sarvam-1' (2B parameter Indic model) and 'Sarvam Edge' for on-device, offline processing. Their suite includes Bulbul (voice), Sarvam Vision (document intelligence), and Akshar, a workbench for high-precision knowledge extraction.
You'll be a good fit if you have
- Proficiency in Python and modern frameworks like PyTorch.
- Strong foundation in deep learning and model debugging.
- Hands-on experience with GPU-based workloads and performance tuning for high-traffic systems.
- Ability to build sophisticated evaluation frameworks for non-deterministic AI outputs.
- Experience with speech, audio, or multimodal systems such as TTS, STT, or voice cloning is a plus.
- Required Experience: 3 to 6 years of relevant industry experience in ML/AI, with a proven track record of deploying models in production environments.
- Bonus Points: Familiarity with large-scale ML infrastructure, cloud platforms (AWS/GCP/Azure), container orchestration (Docker/Kubernetes), and automated MLOps strategies.
Key Responsibilities
- Design and maintain the end-to-end ML systems that power the dubbing platform, including capabilities for transcription, translation, speaker separation, and speech generation.
- Develop and implement automated validation and monitoring frameworks to detect regressions in voice quality, speaker similarity, and timing alignment in production.
- Optimize model inference pipelines to reduce latency and compute costs while ensuring 24/7 reliability, including managing GPU-based workloads and multi-provider routing.
- Build systematic evaluation workflows (both offline and online) to measure performance using metrics such as cosine similarity scores for voice identity and BLEU/METEOR for translation accuracy.
- Run rigorous A/B tests and validation experiments to continuously benchmark Sarvam’s models against global alternatives and ensure measurable improvements in each rollout.
- Collaborate cross-functionally with Product and Engineering teams to integrate specialized sub-agents into a unified workspace.
This is not a cold application
You're in the top 1%. We represent you to Sarvam. Not the other way around.
Sovereign AI Powerhouse
The foundation for India's digital future.
Why Sarvam
AI adoption in India is restricted by models that are primarily trained on Western datasets and English nuances.
Sarvam is solving this by building a 'full-stack' Indian AI platform that respects local languages and cultural contexts, bringing GenAI to the next billion users.
Problems that matter. Company that matters. Infrastructure that matters.
Accelerate economic growth by connecting every talent to the work they are meant to do
Hiring got broken. Too focused on companies. Not enough on people. We're fixing it.
Human judgment plus AI. Talent first.
Built By
We built Grapevine. More than 1 million engineers use it to share real talk about companies, salaries, culture. We saw the problem up close.
Great engineers lost in broken systems.
So we built Round1. Backed by Peak XV and Kae Capital. One mission: connect exceptional engineers to work that matters.