Delivered work
AI video processing & analysis
We recognize objects and situations in video and turn the information you need into input for service decisions.
AI that understands data
and keeps work moving.
Starting from our experience in video and data analytics,
we design ways to use AI that fit your organization’s context.
Video & data: delivered work Language AI: proposed
We bring our analytics experienceto the next stage of AI.
Building on our experience developing video analytics, data analytics, and prediction AI, we propose the next step: work environments that use language models.
Delivered work
VideoSensorsBusiness DB
Processed outputService datasetCleanup · Verification · Connection
Data analysis & processing
We collect scattered data and process it for its purpose, so it becomes information people can read and data other systems can connect to.
- AI data platformWe developed an AI data platform in 2018.
- Data Voucher supplierDesignated as a supplier in 2021 (as of the date designated).
Delivered work
Prediction & IoT AI
We developed AI that uses facility, sensor, and vital-sign data to predict facility output and hazards.
- Fire prevention and hazard prediction AI for industrial facilitiesDeveloping an AI facility management system for fire prevention and hazard prediction, and an equipment control (PLC) system
- Facility output forecasting AIDeveloping an AI algorithm for output forecasting
- Wearable for predicting infant safety risks (IoT)Developing a wearable device that collects vital-sign data and predicts risky situations, and a vital-sign monitoring application
Toward AI that understands language and context
LLMs, RAG, and AI agents are proposed use cases. We first check your organization’s data and security environment, then validate what’s possible on a small scale.
Our role and the key technologies in each video, data, and prediction AI project are listed on our project experience page.View projects by industry
We shape AI into three approachesto fit your business challenges.
Enterprise knowledge search (RAG) and document & workflow automation are proposed use cases. Video AI & data analytics builds on projects KPS has delivered itself.
Three use scenariosin everyday work
These proposed scenarios place enterprise knowledge search (RAG) and document & workflow automation in real work situations. We check your current work and data environment and start with a scope that fits.
Proposed scenarioEnterprise knowledge search (RAG)
AI that finds and answers
from internal knowledge
We propose a knowledge search environment that connects work documents and rules, understands the context of a question, and answers with sources.
- Unified search across documents and rules
- Answers built on sources
- Access based on permissions
Work request
Find the security check items in last quarter’s operating guidelines.
Organization data checkAI analysisHuman review
Answer with sources
Security checks follow this order: confirm access permissions, review the change history, and record anomalies.
- Operating guidelines · Security checks
- Access permission management standards
- Change history review procedure
Discuss AI adoption
Start a consultation Tell us about your current work and data environment, and we’ll work out a feasible starting point with you.
We connect AI securelyto existing systems.
We design everything from service screens to data, AI models, security, and operations as one structure. The setup can vary with your organization’s environment.
Example architecture
Business users · Services
AI orchestrationLinks questions, tools, and workflows
Enterprise data
AI models
Security · Access control · Monitoring · Human review
We start smalland let the results decide the next step.
LLM and RAG adoption starts with a pilot (proof of concept, PoC) of limited scope. We set the evaluation criteria first and use the same criteria to decide whether to scale up.
Question selection
From recurring questions or document tasks, we choose a scope whose results are easy to check.
Evaluation set
We first build the criteria by pairing representative questions with expected results and source documents.
Limited rollout
We connect it for a defined set of users and data, including permissions and review procedures.
Scale-up decision
We look at the evaluation results together with the operating workload to decide the next step.
- Scale up
- Refine
- Stop
Let’s start by exploring what AI can do
for your organization.
Tell us about your work and data environment, and we’ll design a realistic starting point with you.