Who we are
We are a team of experts who produce cutting-edge RLHF and SFT training data for AI models. We lead teams of hundreds of labelers with domain-specific expertise and work together to produce the best RLHF and annotated data for AI model training. We have experts with PhD degrees in over 20 fields and proficient in 30+ languages.
- Anderson Lu - CEO & Founder, Math Lead
- Albert Lu - CTO & IT Lead
- Adi Arora - CS Lead
- Michael Margolin - Statistics & Data Science Lead
- Chang Kyun Ha - Physics Lead
- Josh Mizrahi - Humanities and Arts Lead
- Rohan Vijayan - Biomedical Engineer
- Our Team of Hundreds of Contributors
Anderson Lu, from the University of Maryland, College Park with a major in Computer Science (machine learning track) and a minor in Mathematics, is the founder and CEO of Stokes AI: a human-in-the-loop AI training company focused on producing high-quality RLHF and SFT datasets for large language models. Anderson holds CompTIA Security+ and Network+ certifications, and he has used such knowledge to train AI models. In the past, he had worked extensively with Outlier AI (Scale AI) and DataAnnotation.tech (Surge AI), where he evaluated AI model outputs, ranked responses, crafted prompts to test AI model's capabilities with math reasoning, and provided STEM domain-specific feedback for AI responses. His experience in the RLHF/SFT industry inspired him to take the next step by building his own company—offering expert-crafted training data with an emphasis on STEM reasoning, precision, and diversity of feedback. He also oversees math projects where AI models are trained to think and reason through complicated math problems. His LinkedIn can be found here.
Albert Lu is a seasoned IT consultant with over 20 years of experience delivering technical solutions and operational leadership across both federal agencies and private-sector enterprises. He holds a graduate degree from the Whiting School of Engineering at Johns Hopkins University and has built a distinguished career specializing in Linux system administration, AWS cloud architecture, Cisco network management, high-availability clustering, and cybersecurity compliance frameworks such as NIST and FedRAMP.
At Stokes AI, Albert plays a critical role in transferring his real-world IT expertise into AI systems. He works closely with our data engineering and annotation teams to create high-fidelity training sets that enable AI models to learn complex tasks such as provisioning infrastructure, diagnosing network issues, managing servers, and responding to security threats—mirroring the work of skilled IT professionals. His ability to translate decades of technical know-how into structured, trainable data allows our models to not only understand but also emulate expert-level decision-making in dynamic environments. Albert’s work ensures that the next generation of AI systems are grounded in practical, enterprise-grade knowledge and capable of supporting mission-critical operations at scale. His LinkedIn can be found here.
At Stokes AI, Albert plays a critical role in transferring his real-world IT expertise into AI systems. He works closely with our data engineering and annotation teams to create high-fidelity training sets that enable AI models to learn complex tasks such as provisioning infrastructure, diagnosing network issues, managing servers, and responding to security threats—mirroring the work of skilled IT professionals. His ability to translate decades of technical know-how into structured, trainable data allows our models to not only understand but also emulate expert-level decision-making in dynamic environments. Albert’s work ensures that the next generation of AI systems are grounded in practical, enterprise-grade knowledge and capable of supporting mission-critical operations at scale. His LinkedIn can be found here.
Adi Arora is a rising senior at UC San Diego majoring in Math–Computer Science. He is passionate about using AI and machine learning to enhance the effectiveness of real-world systems. Adi had built a computer adaptive mastery quiz that personalizes and gamifies learning using AI, showcasing his interest in applying ML to education. He aspires to become a machine learning researcher and currently leads an AI-focused club at UCSD, where students explore the history and future of artificial intelligence. Adi is very passionate about pushing the frontier of the machine learning field. Outside of academics, Adi enjoys going to the gym, playing poker with friends, and watching Netflix. His LinkedIn can be found here.
Jordan is a rising senior attending Virginia Tech majoring in Systems Engineering and minoring in Computer Science. His deep interest in financial markets—spanning stocks, crypto, and technical analysis—led him to develop a trading system that capitalizes on market inefficiencies to generate positive expectancy. Jordan currently teaches a cohort of students how to succeed in trading, but his broader vision lies in training AI models with this financial expertise. By translating trading strategies, chart patterns, financial news interpretation, and economic indicators into structured datasets, Jordan helps Stokes AI build models that reason like a seasoned trader—capable of analyzing charts, recognizing signals, and making sound market predictions. His LinkedIn can be found here.
Josh Mizrahi is the humanities Team Lead at Stokes AI, where he oversees projects involving arts and humanities. He also oversees projects relating to translation, interpretation, and culturally nuanced training for large language models. He is currently majoring in Persian Language, Culture, and History at the University of Maryland, with a strong academic focus on comparative anthropology. While his work is centered on Persian anthropology and linguistics, Josh brings a global perspective to the table—drawing on his studies of Middle Eastern, South Asian, and Mediterranean civilizations to ensure AI systems are trained with a deep and balanced understanding of cultural context. With experience in both modern and classical literature, oral traditions, and ethnographic methods, Josh helps ensure that AI models reflect the richness and diversity of human communication. He is particularly passionate about preserving linguistic nuance and historical depth in multilingual AI systems, making him an essential bridge between humanistic scholarship and cutting-edge machine learning. His LinkedIn can be found here.
Michael serves as the Statistics and Data Science Lead at Stokes AI, where he oversees and manages AI training projects rooted in statistical analysis and data-driven reasoning. He holds an M.S. in Applied Statistics from Penn State University and a B.A. in Applied Mathematics from Queens College, CUNY, and has successfully completed the Society of Actuaries exams in Probability and Financial Mathematics. Michael possesses robust expertise in Python, R, SAS, and advanced statistical modeling, covering regression analysis, hypothesis testing, and machine learning. His recent experience includes contributing to Snorkel AI, where he designed original mathematics problems and rigorously evaluated large language model outputs for mathematical logic and statistical accuracy. At Stokes AI, Michael leverages his deep background in data science to develop and validate challenging reasoning datasets, ensuring AI systems can tackle complex statistical problems and support rigorous quantitative research. His commitment to enhancing AI capabilities in mathematics and statistics positions him as a vital bridge between mathematical theory, applied data science, and next-generation intelligent systems.
His project portfolio is available on GitHub and he can be found professionally on LinkedIn.
Chang Kyun Ha is the Physics Lead at Stokes AI, bringing over seven years of combined academic and industry experience in photonics and optical science. With a Ph.D. in Physics from the Korea Advanced Institute of Science and Technology (KAIST), Dr. Ha has conducted advanced research in nonlinear optics and nanophotonics, specializing in pulsed laser systems, quantum communication, and structured light. His industrial contributions span mobile lens system design at Samsung Electro-Mechanics to pioneering femtosecond laser processing for glass substrates at LG Innotek. Leveraging a portfolio of publications and deep technical expertise, Dr. Ha channels his passion for physics into the development and training of AI systems capable of solving complex physics problems, assisting with experimental design, and reasoning through advanced scientific challenges. By blending rigorous scientific knowledge with AI training, he helps ensure that Stokes AI’s models can support and accelerate innovation in both academia and industry. His LinkedIn can be found here.
Rohan Vijayan, Ph.D., is a Staff Software Engineer at Stryker and a distinguished contributor to Stokes AI, specializing in medical imaging, surgical robotics, and computer vision. He holds a Doctorate in Biomedical Engineering from Johns Hopkins University, with additional degrees in Biomedical Engineering, Computer Science, and Scientific Computing from Vanderbilt University.
With a robust technical background in Python and extensive experience in both academic and industry settings, Dr. Vijayan has led multidisciplinary projects involving prototype system design and performance validation across phantom, cadaver, and animal studies. His work is supported by multiple peer-reviewed publications, conference presentations, and patent applications.
At Stokes AI, Dr. Vijayan uses his experience in computational medicine and biomedical engineering to design and train advanced AI systems for healthcare applications. His expertise drives the development of models capable of complex reasoning about medical imaging, surgical planning, and robotic intervention, ensuring that AI technologies reflect the highest standards of clinical rigor and innovation. Fluent in both technical development and applied research, he represents a vital link between human expertise and machine learning in the future of medical technology. His LinkedIn can be found here.
At Stokes AI, Dr. Vijayan uses his experience in computational medicine and biomedical engineering to design and train advanced AI systems for healthcare applications. His expertise drives the development of models capable of complex reasoning about medical imaging, surgical planning, and robotic intervention, ensuring that AI technologies reflect the highest standards of clinical rigor and innovation. Fluent in both technical development and applied research, he represents a vital link between human expertise and machine learning in the future of medical technology. His LinkedIn can be found here.
[bio3] [linkedin3]
[bio4] [linkedin4]
[bio5] [linkedin5]
Behind every dataset at Stokes AI is a team of hundreds of specialized contributors—each an expert in their own field. From linguists and subject matter experts to experienced annotators and QA reviewers, our network ensures that every annotation is context-aware, accurate, and relevant. This multi-disciplinary approach allows us to deliver training datasets that not only meet technical standards but also reflect real-world diversity, nuance, and depth.
About our name
Stokes AI is named in honor of Sir George Gabriel Stokes (1819–1903), the renowned Irish mathematician and physicist whose foundational work in fluid dynamics, optics, and vector calculus continues to influence modern science and engineering. His name lives on in the Navier–Stokes equations, which describe the motion of fluids, and Stokes’ Theorem, a cornerstone of multivariable calculus used in fields from electromagnetism to differential geometry. He also did research into the absorption and scattering of light, which provided a basis for spectroscopy.
While artificial intelligence did not exist during his time, his legacy is deeply embedded in the mathematical structures and physical principles that many advanced AI systems rely on today. His work underpins areas including physics-informed machine learning, computational fluid dynamics, computer vision, and geometric deep learning—domains where AI increasingly plays a role in solving complex, real-world problems.
We chose the name Stokes to reflect our commitment to rigor, clarity, and the seamless integration of data with domain knowledge. Just as George Stokes bridged abstract mathematics with natural phenomena, we aim to build AI systems that are not only powerful, but interpretable, principled, and grounded in the systems they seek to understand.
Contact us
Interested in working together? Fill out some info and we will be in touch shortly. We can’t wait to hear from you!