Research and evaluation services for frontier AI.

Clement is a former Research Engineer at the Singapore AI Safety Institute (NTU DTC) where he worked on the safety of multimodal models and computer-use agents. He has published multiple first-author papers across ICLR, EMNLP and ACL. He is a Research Advisor at Apart Research, where he has advised dozens of projects on technical AI safety over the past three years.

Miro has a decade of experience across Europe and Asia as a product, governance and research management lead with SAP, ByteDance, University of Oxford and GovAI. He is the founder of the Singapore AI Safety Hub and a graduate of Schwarzman College at Tsinghua University.

Isaac led Neo Research’s independent evaluation of Kimi K3 for misalignment risks. He was previously a research intern at A*STAR’s Institute of High Performance Computing, where his work on physics-informed neural networks was published at IJCNN 2026, and a machine learning engineering intern at Synapxe, Singapore’s national HealthTech agency, where he worked on multimodal retrieval for healthcare. At the National University of Singapore, he taught CS1010S for nearly three years, serving as Head Teaching Assistant in his final semester.

Farrel works on agentic and computer-use agent evaluations. He was previously a Member of Technical Staff at HUD (YC W25), where he authored evaluations and reinforcement learning environments for frontier AI labs and helped build an LLM evaluation platform that has powered over one million agent evaluation runs. Before HUD, he spent nearly three years as a Software Engineer at RemNote, where he led development of the PDF and web reader, a feature cited by 30% of paying subscribers as their reason for subscribing, and built LLM pipelines for flashcard generation serving over one million users.

Andreas has over seven years of research experience, with 17 peer-reviewed publications and more than 450 citations. Before joining Neo Research, he conducted independent technical AI safety research under a BlueDot Impact career transition grant, examining whether safety interventions in agentic systems persist under fine-tuning, distribution shift and adversarial adaptation. He was an AI Safety Research Fellow at Safe AI Germany, working on countermeasures to emergent misalignment in open-weight LLMs. Previously, he was Director of AI/ML Engineering at Syndigo and a researcher at the University of Münster, where he managed an EU-funded project spanning six countries.

Neil is a machine learning researcher whose work on data deletion, conducted at the National University of Singapore, was published at ICML and has been cited over 100 times; his more recent work on consistency training was accepted to EMNLP Findings. As an AI Research Fellow at LASR Labs, he studied how imperceptible tokens in benign documents can manipulate LLM decision-making. Before moving into alignment research, he spent four years as a Senior Technical Product Manager for AI at Workato, where he launched agent orchestration, enterprise MCP and document-processing products that together generated over $20M in revenue.

Rhea works on multi-agent propensity evaluations for frontier models. She has over six years of software engineering experience at leading technology companies, most recently as a Senior Software Engineer at Skydio, where she owned connectivity and embedded attachment work for its flight-controller mobile experience and led initiatives in Android platform integration, over-the-air updates and performance optimization. Before Skydio, she was a Software Engineer on Instagram’s Camera AR team.