Start: 3/01
End: 6/31
WatonGroup
LLM 0 to 1
Explored LLM principles and implementation paths from zero — as LLM R&D lead, I drove the architecture experiments.
Key stakeholders: AI strategy committee and data governance team
- Built an end-to-end pipeline spanning tokenizer, data processing, and training
- Established model evaluation and tuning practices, laying the groundwork for future proprietary models
- Defined data collection, cleaning, and compliance approval processes
- Set up training resource management and experiment tracking conventions