Iacon Autonomics develops Optimus, an autonomous reinforcement learning engineer designed to help teams train and improve AI agents. Users provide an agent or model and a goal; Optimus researches algorithms, designs and runs experiments, evaluates results, and decides what to try next. The platform supports work across language models, robotics, diffusion models, game characters, biological systems, and custom policies.

Optimus coordinates a complete training workflow, including data preparation, reinforcement-learning environments, reward-based practice, fine-tuning, preference learning, benchmarking, and evaluation. Teams define environments, rewards, and success criteria, while Optimus runs experiments within their technology stack and promotes model checkpoints when they meet evaluation checks. Iacon says this approach can use established research methods or explore new techniques to produce more useful training data and improve policies.

The company also collaborates with research and technology partners, including compute, simulator, and data providers. Its published case studies cover investigative-agent evaluation, teaching a language model to paint with code, and a robotics testbed.