There are 3 more teams in the continual learning space. Trajectory (http://trajectory.ai) Founded by Ronak Malde (previously worked on post-training at Windsurf, with a team background including DeepMind), Michael Elabd, and Arjun Karanam (former Apple researcher). Funding: In May 2026, raised approximately $15M in a Seed round led by Conviction, with participation from Bessemer, Radical, and individual investors like Jeff Dean and Fei-Fei Li. In August of the same year, raised $40M in a Series A led by Sequoia, with NVIDIA and Bessemer participating, bringing the post-money valuation to around $300M and total funding to approximately $55M. Their business focuses on transforming production trajectories/user corrections into continual post-training and has open-sourced the C-LoRA multi-tenant training stack. C-LoRA: Multi-tenant setup that runs multiple LoRA RL experiments in parallel on the same base. Claims an end-to-end throughput of approximately 2.81× with no reward regression. Includes architecture, concurrent experiments, and SkyRL setup steps. https://trajectory.ai/field-notes/multi-lora-training-for-continual-learning https://(github.com)/NovaSky-AI/SkyRL Baseten / Parsed Parsed is a small team working on reinforcement learning, post-training, and continual learning. Their Chief Scientist, Charles O’Neill, co-authored the research paper on rank-1 LoRA continual learning. In December 2025, they were acquired by the inference platform Baseten. Baseten itself is an inference company founded in 2019, with clients including Cursor, Harvey, and Decagon. It has raised significantly more funding: $300M in January 2026 at a $5B valuation, followed by a $1.5B Series F in June 2026 at a $13B valuation. The acquisition of Parsed was aimed at integrating inference with continual post-training. Thinking Machines Lab A lab founded by OpenAI’s former CTO, focusing on RL + LoRA research. https://thinkingmachines.ai/blog/lora/