AI Engineer to Optimize Existing LLM Performance and more

I already have a large-language model in production, trained solely on text data, and it’s doing well—but “well” isn’t good enough for the scale I’m moving toward. I’m looking for an AI engineer who can dive into the current codebase and training pipeline, diagnose the accuracy bottlenecks, and then fine-tune or refactor the model so that its answers are both more precise and noticeably faster.

The two clear targets are:
• Enhancing accuracy (fewer hallucinations, higher BLEU/F1 on our internal evaluation set)
• Reducing response time (lower end-to-end latency during inference).

I’ll grant you access to the existing checkpoints, tokenizer, evaluation harness, and a labeled text corpus. You’re free to propose techniques such as additional domain-specific fine-tuning, model pruning, quantization, knowledge-distillation, optimized batching, or even a revamped serving stack—whatever achieves the goals without sacrificing stability.

Deliverables I need from you:
Can be discussed. consider this as a pilot project to test the Human intelligence for AI capability :)

If this collaboration works well, there’s scope for ongoing iteration and feature expansion down the road.

I will focus on relevant experience in AI and LLMs rather than past projects.