ACTIVE
Enterprise LLM Application Development
Agents, RAG knowledge bases, mini-programs and on-premise deployment — bringing LLM capability into your business workflows
RAG knowledge baseCustom agents
Scenario · Enterprise AI
Enterprise LLM Applications
Document / contract / resume analysis, knowledge-base Q&A, business agents and AIoT analytics — compliance built in
Custom LLM applications for enterprises, built with the engineering methods from our two in-house AI SaaS products: from requirements and data assessment, through prototype and evaluation set, to build, integration and live operations. Domestic models and deployment by default, meeting in-country data, informed-consent and AIGC labelling requirements.
Problems to solve
- 01The demo looks great, but results are unstable in production — there is no evaluation framework
- 02Unclear data and compliance boundaries
- 03Models and costs change quickly, so switching has to be possible
Architecture
- 01Application
- Web / mini-program / enterprise system integration
- 02Orchestration
- RAG retrieval, agent tool calling, streamed output
- 03Models
- Multi-provider abstraction, domestic first, switchable
- 04Compliance & quality
- De-identification, encryption, retention policy, evaluation regression
What we handle
- Full-stack development, evaluation framework, deployment and operations
What you bring
- Business scenario, sample data and acceptance criteria
Delivery process
- 011 wk
Requirements & data
Scenario, data sources, compliance boundaries, acceptance criteria
- 021–2 wks
Prototype & eval set
Runnable prototype + test cases
- 034–8 wks
Build & integrate
RAG / agent / frontend / system integration
- 04Ongoing
Launch & operate
Staged rollout, monitoring, evaluation regression, model switching
FAQ
Does data from AI applications leave China?
By default we use models hosted in China and deploy in China, so data does not leave the country. If an overseas model is required, the customer makes that choice explicitly after a compliance assessment.
How are IP and confidentiality handled?
We can sign an NDA. Ownership of the design deliverables from a custom project is set out explicitly in the contract, and the customer’s own designs and materials are used only for that project.
Products used
ACTIVE
Aideed · AI Contract Risk Analysis
Upload a contract and see its risks in minutes: itemised issues, four risk levels, conversational follow-up questions
Contract analysisCustom agents
ACTIVE
GetCallback · AI Resume Diagnosis
Understand why you did not get a callback: reasons and fixes, not a vanity score; plus application tracking and cohort benchmarks
Resume diagnosisCustom agents
Related cases
- Legal tech / enterprise servicesIn-house: AI Contract Risk Analysis SaaSAn AI contract-analysis product taken from zero to launch: NestJS modular monolith + Next.js, encrypted contract storage, de-identification before the model, domestic models first; ICP filing completed and live.
- Job-seeking services / consumer AIIn-house: AI Resume Diagnosis SaaSResume diagnosis with no vanity score: reasons and fixes, application tracking and k-anonymous cohort benchmarks, PIPL compliance and AIGC labelling built in, deployed on an Alibaba Cloud ECS shared with other sites.