주식회사 솔트룩스

Climate Impact & Sustainability Data (2021)

Reduced energy consumption by up to 12 times through core technology R&D, resulting in a reduction of CO2 emissions to less than 1/10 of the total.
Improved the performance of the Soinet inference engine by more than 3 times, reducing energy consumption to less than 1/3 through algorithm changes in the Mask R-CNN deep learning model acceleration project.
Achieved a 2x increase in inference speed and 1/4 reduction in computing power for the Google BERT model by applying Kernel Fusion and FP16 operations.

Continued the '1% Sharing Movement' social contribution activity, accumulating over 500 million won in funds for supporting underprivileged groups.
Built and donated over 10 'Haedami' libraries to local children's centers.
Actively participated in establishing principles for a user-centric intelligent information society.

Maintained a board of directors with 50% (2 out of 6) independent directors.
Implemented policies for shareholder-friendly practices, internal transaction control, and internal accounting management.
Regularly conducts IR activities for domestic and international institutional investors.

High GPU memory allocation and slow FPS in existing deep learning models used for projects like the Digital New Deal AI learning data construction project.
Potential for biased data in large-scale data learning, as seen in the 'Iruda' incident.

Collaborated with universities through open innovation to accelerate and optimize deep learning neural networks for disaster waste detection.
Assigned an internal executive to oversee the personal information protection of AI learning data, implementing measures such as access restrictions, encryption, and masking.
Proactively addressed data bias concerns through verification and ensuring copyright and license compliance for learning data.

Climate Change Data