- Bengaluru, Karnataka, India
- Full-time
- ₹3M–₹4M per year
- 3+ years
Design, develop, and scale advanced pipelines for generating, executing, and validating synthetic training data for safety-critical embedded systems.
Compensation
Salary range: Rs 2500000 - Rs 3500000 (ie INR 25 - 35 LPA)
Responsibilities
- Design and develop scalable pipelines to generate, execute, and validate synthetic training data, including code, technical specifications, requirements, and other structured technical documents.
- Build automated validation systems using compilers, test harnesses, execution environments, and formal verification tools rather than relying solely on LLM-based evaluation.
- Develop and maintain reward models, verifiers, and evaluation mechanisms for reinforcement learning post-training approaches such as Group Relative Policy Optimization (GRPO).
- Own end-to-end dataset quality, including data decontamination, filtering, coverage analysis, validation, and comprehensive documentation.
- Build and maintain synthetic data generation and instruction-tuning pipelines for code and structured technical content.