Elite LLM Engineer / AI Team — Deep LLM Expertise & Harness Building

We are building our own large language model, and I'm looking for an elite AI engineering team (or an exceptional individual who can lead one) with genuinely deep LLM expertise. What we're doing: Training and fine-tuning our own LLM, and building the harnesses, scaffolding, and evaluation infrastructure that let the model actually perform in real products. What you'll be building: LLM harnesses and agent scaffolding (tool use, function calling, memory, retrieval, multi-step reasoning loops), training and fine-tuning pipelines (pre-training, SFT, RLHF/DPO), evaluation harnesses and benchmarking suites, inference serving and optimization (quantization, batching, caching, latency and cost tuning), and the data pipelines behind all of it. What I need from you: A deep, first-principles understanding of how LLMs actually work (tokenization, attention, context windows, sampling, alignment) rather than just API-level familiarity. Proven experience building harnesses for LLMs and agentic systems. A strong software engineering foundation: Python, PyTorch, distributed training, clean and testable code, and solid architecture instincts. Experience shipping AI systems to production, not just prototypes or demos. Someone who can brainstorm, challenge assumptions, and make technical decisions alongside me. Nice to have: Open-source contributions to LLM tooling, experience with vLLM / Ray / DeepSpeed / Hugging Face, custom eval frameworks, and multi-model or multi-provider orchestration. Engagement: Full-time, long-term, more than 30 hrs/week. Please start your proposal by describing a specific LLM harness or eval framework you built, what the hardest technical problem was, and how you solved it. Generic or templated proposals will be ignored. Timeline: Looking to start within the next 1-2 weeks.