Yazi L.

I build and fine-tune AI systems that solve real optimization and engineering problems, from LLM agents to reinforcement learning pipelines. My background combines a Computer Systems Engineering degree with hands-on research in applying large language models to hard technical domains, including a graduation research project at NYU Abu Dhabi where I developed an LLM-guided system to optimize compilers for encrypted computation. What I bring to a project: Fine-tuning and training LLMs with reinforcement learning (GRPO, PPO) and supervised fine-tuning, including reward design and evaluation Building agentic LLM pipelines and tool-using AI systems for automation and decision-making tasks Applying evolutionary and search-based optimization methods (MAP-Elites) alongside neural approaches Machine learning model development end to end: data preparation, training, evaluation, deployment Strong Python engineering background, comfortable working across the full stack when a project needs it (Django, React, Node.js) Experience running large-scale training jobs on HPC clusters (SLURM, multi-GPU setups) and debugging the messy realities of that (CUDA issues, OOM errors, tokenizer bugs) Recent work includes building a skill-matching recommendation engine using fine-tuned language models and vector search, reaching 87% accuracy in production use, and developing CHEHAB, a research system combining agentic LLMs and evolutionary search to automate compiler optimization for privacy-preserving computation. I care about writing code that works reliably, not just in a demo. I communicate clearly and regularly, ask questions early rather than guessing, and deliver work that's documented well enough that you're not stuck depending on me to maintain it. If your project involves LLM fine-tuning, AI agents, machine learning model development, or general applied AI engineering, I'd like to help.