Carrierbot Develops Job-Hunting-Specific LLM Based on 'Gemma 4'. Promoting Secure AI Use in Universities and Government via On-Premise Support.
Carrierbot has developed 'shukatsu-gemma4', an LLM specifically for job-hunting support based on Gemma 4, addressing security and dialogue quality issues in educational settings. It has been released on Hugging Face for local execution.
📋 Article Processing Timeline
- 📰 Published: April 8, 2026 at 03:10
- 🔍 Collected: April 7, 2026 at 18:30
- 🤖 AI Analyzed: April 20, 2026 at 21:19 (314h 49m after Collected)
## ■ Background of Development: Unique Challenges in Educational Settings
As various generative AIs are developed, there have been primarily two challenges in employment support settings such as universities.
Security Risks: Concerns about transmitting students' sensitive personal information to external clouds.
The Barrier of Dialogue Quality: The fact that general-purpose AI cannot replicate the complex context specific to new graduate recruitment or the tone of an advisor.
## ■ Technical Features and Evaluation: Balancing Local Operation and Practical Quality
By strictly selecting and learning only high-quality dialogues from over 10,000 job consultation records, this model achieves structural resume (ES) corrections and 'empathetic dialogue' tailored to practical work. In comparative verification with major LLMs, while it yields to large-scale cloud AIs in comprehensive knowledge volume, it has confirmed an equivalent level of evaluation in the 'quality of ES correction' and 'dialogue tone'. It is a highly practical model ready for immediate deployment in the field, despite being in a local environment that protects confidential information.
## ■ Published on 'Hugging Face' Aiming for Industry-Academia-Government Co-creation
To elevate Japan's employment support infrastructure and accelerate 'co-creation' by industry, academia, and government, we are publishing this model on 'Hugging Face'.
Repository Name: careerbot/shukatsu-gemma4
Ease of Implementation: Adopts 'Gated Access' which allows immediate access simply by agreeing to the terms of use. By utilizing tools like Ollama and LM Studio, it is possible to verify and deploy a highly accurate job support AI in a local environment.
*This model is a fine-tuned version of Google's 'Gemma 4' and is not an official Google product.
## ■ Experiential Exhibition at 'GENIAC' Sponsored by METI and NEDO
This model will be demonstrated with actual machines at the result presentation caravan of 'GENIAC', a generative AI project led by the Ministry of Economy, Trade and Industry (METI) and NEDO. Actual operations...
As various generative AIs are developed, there have been primarily two challenges in employment support settings such as universities.
Security Risks: Concerns about transmitting students' sensitive personal information to external clouds.
The Barrier of Dialogue Quality: The fact that general-purpose AI cannot replicate the complex context specific to new graduate recruitment or the tone of an advisor.
## ■ Technical Features and Evaluation: Balancing Local Operation and Practical Quality
By strictly selecting and learning only high-quality dialogues from over 10,000 job consultation records, this model achieves structural resume (ES) corrections and 'empathetic dialogue' tailored to practical work. In comparative verification with major LLMs, while it yields to large-scale cloud AIs in comprehensive knowledge volume, it has confirmed an equivalent level of evaluation in the 'quality of ES correction' and 'dialogue tone'. It is a highly practical model ready for immediate deployment in the field, despite being in a local environment that protects confidential information.
## ■ Published on 'Hugging Face' Aiming for Industry-Academia-Government Co-creation
To elevate Japan's employment support infrastructure and accelerate 'co-creation' by industry, academia, and government, we are publishing this model on 'Hugging Face'.
Repository Name: careerbot/shukatsu-gemma4
Ease of Implementation: Adopts 'Gated Access' which allows immediate access simply by agreeing to the terms of use. By utilizing tools like Ollama and LM Studio, it is possible to verify and deploy a highly accurate job support AI in a local environment.
*This model is a fine-tuned version of Google's 'Gemma 4' and is not an official Google product.
## ■ Experiential Exhibition at 'GENIAC' Sponsored by METI and NEDO
This model will be demonstrated with actual machines at the result presentation caravan of 'GENIAC', a generative AI project led by the Ministry of Economy, Trade and Industry (METI) and NEDO. Actual operations...