- Model adaptation & training
- Task-specific benchmarks focused on functional/executable correctness
- Standing up/optimizing infrastructure
- Reproducible pipelines (versioning, checkpoints, seeds, experiment tracking), structured generation.
Key Responsibilities :
- Design and execute end-to-end training and fine-tuning pipelines for LLMs to improve task-specific performance and domain accuracy.
- Implement advanced alignment techniques including DPO and GRPO to ensure model outputs are secure, reliable, and contextually relevant.
- Optimize model deployment and inference efficiency using techniques like LoRa and QLoRa to reduce computational overhead while maintaining high precision.
- Conduct continuous performance benchmarking and evaluation of CPT and fine-tuned models to ensure they meet rigorous quality standards before production release.
- Collaborate with cross-functional teams to integrate OLLama and Llama-based solutions into existing enterprise platforms, ensuring seamless scalability and performance.