MxM: Building a Smarter, Fairer Future for AI Talent Recruitment

MxM, founded by UCL Mathematics Professor Hao Ni, is redefining AI talent recruitment with its AITalentBench platform, which uses real-world simulation to evaluate and upskill candidates.

As global demand for AI expertise accelerates—an estimated 97 million new roles are expected in the coming years—traditional hiring methods such as CV screening and technical interviews are proving increasingly inadequate. MxM is developing AITalentBench as a scalable, transparent, and evidence-based platform that transforms how companies identify, assess, and develop AI professionals. 

During cohort seven of the CDI Impact Accelerator, the MxM team advanced this vision by developing a minimum viable product (MVP), refining their CAPSULE simulation engine, and building the cloud infrastructure to support a truly global AI talent pipeline. 

The Challenge: Time-Consuming, Biased, and Inefficient Recruitment

Hiring for AI and data science roles remains one of the most inefficient and error-prone processes in modern recruitment. On average, it takes around 80 days to fill a single role (Workable, 2023), often involving lengthy interview processes and subjective CV screening that can introduce unconscious bias and overlook genuine talent. Furthermore, Professor Hao Ni highlights that “A wrong hire can be extremely costly—up to twice the salary of the position—and the lack of standardisation in assessing technical skills further complicates the process.” 

For small and medium-sized enterprises, these challenges are even more acute. Many struggle to attract a diverse range of qualified applicants, and the effort required to fairly benchmark technical competence across a large number of candidates is beyond the scope of most recruitment teams.  

“As the World Economic Forum has highlighted, the growing demand for AI, machine learning, and data science professionals far exceeds current supply, creating an urgent need for smarter, fairer, and more scalable methods of talent identification and evaluation”, says MxM founder.  

The Solution: Simulated, Agent-Assisted AI Talent Evaluation 

The solution MxM is developing, AITalentBench, tackles this challenge with a new approach—placing candidates in realistic AI scenarios through its CAPSULE simulator, which generates role-specific tasks powered by large language model (LLM) agents.  

“Instead of relying on generic coding tests or academic quizzes, the platform mirrors real-world problems aligned to job requirements,” says founder Hao, “giving applicants the chance to demonstrate their creativity, problem-solving, and technical decision-making.” 

Each CAPSULE is tailored using automated task creation powered by LLMs, and candidates receive guided support from AI agents throughout the process. These agents offer coaching, suggest baseline solutions, and simulate peer collaboration, mirroring how real teams solve problems.  

The result is a more accurate, holistic picture of a candidate’s skill set. Evaluations are explainable and benchmarked, giving hiring managers transparent insight into how a candidate performed and why.  

Accelerating Development with the UCL CDI

Through the CDI Accelerator, with the expert guidance of AWS Solutions Architects, UCL Advanced Research Computing (ARC) and UCL School of Management, AITalentBench advanced rapidly from concept to market-ready product, hitting key technical and business milestones along the way. 

Technical Acceleration:

The team delivered the MVP of AITalentBench, complete with a live CAPSULE generator and LLM-powered agent workflow for dynamic task creation and problem-solving support. They implemented evaluation pipelines, interactive leaderboards, and a scalable AWS cloud infrastructure—laying the foundation for high-performance delivery and global scalability. 

Business Acceleration:

On the business front, MxM refined its go-to-market strategy with expert guidance and targeted financial services research. Clear subscription tiers were established for recruiters, upskilling providers, and job platforms, ensuring the platform delivers tailored value to multiple market segments. 

The accelerator also enabled the team to complete a research manuscript, "Structured Agentic Workflows for Financial Time-Series Modeling with LLMs and Reflective Feedback", accepted by the International Conference on AI in Finance (ICAIF) 2025—further validating the platform’s innovative, real-world approach. 

Looking Ahead: Scaling the Future of AI Hiring 

With the MVP live, MxM Innovations is enhancing AITalentBench’s CAPSULE engine with retrieval-augmented generation (RAG) and human-in-the-loop oversight to ensure task quality and fairness. Pilot deployments with financial institutions will test the platform in real-world recruitment scenarios. 

On the business side, the team is pursuing seed funding to scale infrastructure and accelerate go-to-market efforts, while building partnerships with universities, recruitment platforms, and enterprise clients. 

By merging cutting-edge AI with thoughtful design and real-world relevance, MxM Innovations is addressing one of the most urgent bottlenecks in the future of work—helping companies and candidates meet in a smarter, fairer hiring ecosystem. 

To learn more about AITalentBench please contact: Professor Hao Ni - h.ni@ucl.ac.uk

Acknowledgements: 

Support provided to MxM throughout the CDI Impact Accelerator programme included technical expertise from Tomasz Grzybowski (AWS Solutions Architect), Igor Tseyzer and Socrates Varakliotis (UCL Advanced Research Computing); business and strategic guidance from Magda Hercheui (UCL School of Management); and ongoing programme support from the wider CDI team. 

References: 

Workable. 2023. 'Key hiring metrics: Useful benchmarks for tech roles'. Available at: https://resources.workable.com/stories-and-insights/key-hiring-metrics-for-tech-industry