Principal AI/ML Agentic Systems Engineer

Arthur Colle possesses deep knowledge of OpenAI, Anthropic, and LLM APIs, capable of running models on GPUs and Apple Silicon unified memory. He has trained and deployed agentic coding, tool calling, and multi-turn agent models in two prior roles. Currently, he runs a small R&D shop (Distributed, Systems & Co.) focused on distributed AGI and AI safety research, developing graphsub, a graph-native service mesh for agents, eval harnesses, and self-modifying DSLs. Prior to this, he spent 4 years at Goldman Sachs working on SecDB/Slang, collateralized mortgage obligations, interest rate products, and CDS lifecycle management, earning a promotion from tech into structuring and holding a Series 7. Recent work includes building a 15-service LLM stack (zero downtime across 98 deploys), fine-tuning Llama-3.1-8B (23% hallucination reduction), and developing Rust daemons and TUIs, with over 3k GitHub contributions last year. He has contributed to OSS projects like agents.erl, openai-mcp, fortitude, and dspy.ex. Arthur is seeking IC or research roles in AI labs, defense tech, or startups building agentic systems, foundation model infrastructure, or AI safety tooling. He is proficient in running SFT and RL experiments, training models, and is eager to train agentic models with large GPU clusters, having focused on building 'agentic scaffolding' and scaling multi-agent systems for the past two years. He has also published recent research.