Multi-face intelligence
Tracks independent face regions and applies localized enhancement instead of whole-frame filtering.
APPLIED AI · B2B / B2C / EMBEDDED
AI solutions built to move from model to market.
JunleeBuild turns AI architecture into focused products for businesses and the people they serve—from real-time visual intelligence to private workflows and multilingual experiences.
Real-time & cloud AI Human-governed systems Deployable product architecture
LIVEFIT AI · Visual Intelligence
Face-aware real-time media pipeline
FACE REGION · 01 SEMANTIC MASK
Processing On-device frames remain local
Runtime target Windows RTX OBS integration path
Acceleration GPU path Maxine + CUDA
NVIDIA Maxine integrationface box + landmark path
Local semantic maskingskin · hair · clothing separation
OWN IP
repeatable product systems, not outsourced projects
B2B2C
software for teams, platforms and end users
EDGE + CLOUD
workload-aware inference and delivery
HUMAN-GOVERNED
explicit confirmation for consequential actions
FLAGSHIP VISUAL AI
Technical Beta · Windows RTXReal-time face-aware video processing for creator workflows.
LIVEFIT AI is an OBS filter engine designed for RTX-class production. Its source integrates NVIDIA Maxine AR SDK, CUDA and Direct3D 11 paths for multi-face tracking, localized enhancement and fail-safe passthrough.
Designed for
Creator softwareLive productionVideo & commerce platforms
Privacy-aware processing
Core video analysis is designed to run on the user’s device; optional external features stay separately controlled.
Fail-safe media path
Detection loss, unsupported graphics paths and repeated GPU errors are designed to preserve the original frame.
Integration path
NVIDIA Maxine AR SDKCUDADirect3D 11ONNX RuntimeOBS Studio
Source and portable tests are verified. Windows RTX runtime, long-session performance and signed public distribution remain in validation.
NVIDIA, GeForce RTX and NVIDIA Maxine are trademarks or registered trademarks of NVIDIA Corporation. No partnership, endorsement or NVIDIA Inception membership is claimed. Legal notices
PRODUCT & BUSINESS MODEL
One product engine. Multiple paths to market.
We build reusable AI systems once, then package them as business software, end-user subscriptions or embedded customer experiences.
B2B solutions
Pilotable software and integrations for teams with a defined operational problem.
B2C subscriptions
Focused AI products with clear user value and an upgrade path from beta to subscription.
B2B2C embedded AI
Product modules that platforms can deliver to their own creators, customers or members.
Productized architecture
Shared components shorten the path from a validated workflow to a repeatable product.
Privacy-aware boundaries
Local processing, data minimization and explicit provider boundaries are selected by workload.
Human-governed automation
High-impact actions remain visible, reviewable and approval-gated.
TECHNOLOGY ARCHITECTURE
The right inference path for each product.
Latency-sensitive visual work stays close to the device. Data-rich product workflows use portable application and cloud boundaries. Operations remain observable and human-controlled.
Edge inferenceGPU-aware runtimeLocal modelsTyped APIsPortable containersMicrosoft AzureAzure AIPostgreSQLHuman approvalEvidence records
Edge & GPU runtime
Real-time media paths combine hardware-aware inference, local models and native rendering where latency matters.
NVIDIA Maxine pathCUDADirect3D 11ONNX Runtime
Application intelligence
Typed APIs and model adapters turn focused AI capabilities into usable product workflows.
TypeScriptNode.jsPythonModel adapters
Data & retrieval
Structured records, object storage, search and local-first options are composed around the product’s privacy needs.
PostgreSQLObject storageSearchLocal models
Delivery & governance
Reproducible builds, managed runtimes, evidence records and explicit approvals support safe iteration.
DockerManaged runtimesHuman approvalObservability
SELECTED SOLUTIONS
A focused portfolio, backed by working systems.
Each solution is presented with its current readiness boundary. Product availability and external validation differ by track.
Visual & creator AI
Real-time visual processing and browser-first production tools for creators and media platforms.
Customer intelligence
AI-assisted analysis that turns customer feedback into operational priorities and response drafts.
Private workflow AI
On-device document assistance with local models and explicit confirmation boundaries.
Language products
Mobile-first learning and multilingual interaction workflows for real travel and work scenarios.
READINESS YOU CAN AUDIT
Built, verified and still-to-prove are different states.
We keep source implementation, local runtime evidence, external provider validation and public release readiness separate.
Evidence-led product development
STATIC UI · NO LIVE CUSTOMER DATA
Flagship LIVEFIT AI visual AI technical beta
B2B product ReviewPulse local product runtime verified
Private AI Local Doc AI core workflow verified
SOLUTIONDELIVERYEVIDENCESTATUS
LIVEFIT AI Windows RTX Source + portable tests RTX validation
ReviewPulse B2B SaaS Local production runtime Provider validation
Local Doc AI On-device PDF + local LLM workflow Local beta
AI Nihongo Consumer app Production build + runtime AI/Speech E2E
Product release proposal
Illustrative human-governed workflow
1
Collect evidence Build, runtime and product boundary recorded
2
Generate proposal Exact scope and release candidate identified
3
Verify current state Source and environment must still match
4
Execute approved release Blocked until explicit approval
Default boundaries
Destructive actions Denied
Secret collection Denied
Unapproved release Denied
Read-only diagnostics Allowed
Evidence levels
BUILD
Source and tests support the capability
Implementation is not a live customer deployment
LOCAL
A production-like local runtime was exercised
Provider and device boundaries remain explicit
EXTERNAL
Hardware or provider behavior was observed
Only the exact tested path is described
RELEASE
Signed, deployed and supportable product evidence
Public availability is claimed only here
Claims follow evidence
This keeps technical ambition high without blurring what customers can use today.
COMPANY & FOUNDER
An AI product company built around reusable systems.
JunleeBuild is a Seoul-based technology company that designs its own applied AI products and the shared architecture required to operate them.
JL
Yeonjun Lee
Founder & CEO · AI product architecture and full-stack development
- LEGAL ENTITY
- JunleeBuild Co., Ltd.
- ENTITY TYPE
- Corporation registered in the Republic of Korea
- HEADQUARTERS
- A705, Room 402, 4F, 20 Hakdong-ro 24-gil, Gangnam-gu, Seoul, Republic of Korea (Nonhyeon-dong)
- INCORPORATED
- February 10, 2026
- FOCUS
- Applied AI products · B2B2C
PILOTS & PARTNERSHIPS
Bring us a real workflow, not a vague AI brief.
For product pilots, platform integrations, creator technology and technical due diligence.
admin@junleebuild.com