LANGX AI English Speaking Practice & CEFR Assessment

Equmenopolis, Inc. is a Waseda University spin-off established in May 2022 and headquartered in Nishi-Waseda, Shinjuku, Tokyo, led by CEO Yoichi Matsuyama. It builds conversational AI agents from Waseda research and applies them to a hard problem in language education: in a class of 40, students rarely get enough speaking time, and spoken interaction is difficult to assess without human experts.

LANGX for schools links three functions. Practice gives each student goal-based (CAN-DO) conversations with an AI that adjusts level from Pre-A1 to C2 in real time, waits or offers hints using camera-based cues, and quotes the student's own utterances in feedback. Assessment uses a patented engine to rate spoken interaction on the CEFR, with a three-digit score and six components, returning reports within 30 minutes. An analytics tool shows teachers and boards of education what each student can and cannot yet do. The technology rests on the InteLLA dialogue agent and the Gnowsis evaluation engine, and a speech recognition model tuned to Japanese learners.

• CAN-DO practice: goal-based AI conversations aligned with textbook units and scaffolded grammar and vocabulary drills.
• Real-time level adjustment: the AI estimates speaking level from Pre-A1 to C2 and adapts topics and speed.
• Multimodal dialogue: camera cues help the AI decide whether to wait, prompt or give hint phrases.
• Patented CEFR assessment: seven levels, a three-digit score and six components including interaction.
• Teacher analytics: practice counts, speaking time, achieved CAN-DOs and AI-generated reports with searchable utterances.
• Learner-tuned speech recognition: a proprietary model optimised for Japanese learners' pronunciation.

• Provincial departments of education in Vietnam piloting AI speaking practice aligned with textbook units in lower secondary schools.
• Universities using CEFR speaking assessment for placement and graduation requirements.
• Private school chains measuring speaking progress across campuses with class- and school-level reports.

• Speaking time for every student: each learner gets individual AI conversation practice even in large classes.
• Expert-level measurement: CEFR-based assessment of spoken interaction with results in under 30 minutes.
• Evidence for decision-makers: results aggregate by individual, class, school and municipality to measure the effect of AI-supported teaching.
• Proven scale: more than 10,000 speakers use LANGX each year at Waseda University.

Ministries of education across Asia are putting more weight on speaking and interaction, but teachers lack time and tools to assess them. Purpose-built AI tutors that follow curriculum goals, rather than general chatbots, are gaining public-sector budgets. Vietnam's policy to make English a second language in schools from 2025 to 2035 creates similar demand for scalable speaking practice and standardised assessment.

Established May 2022 as a Waseda University spin-off; HQ Nishi-Waseda, Shinjuku, Tokyo; CEO Yoichi Matsuyama. Investors include JST (SUCCESS programme); raised 250 million yen in total in a pre-Series A round in March 2025. Awards include Software Japan Award (2026), Telecom Interdisciplinary Research Award (2026) and SIGDIAL 2024 Best Paper. Users include Kochi Prefecture and Ube City boards of education and Waseda University; also offers LANGX for Eiken.