EPITHRE.

Five AI products, one stack, one building.

Epithre is an independent AI company headquartered in Jakarta. We operate the full AI stack on owned infrastructure in Indonesia, from inference runtime to consumer applications and a developer platform.

Products

Our products.

Slate
Research workspace

Academic search across arXiv, Semantic Scholar, PubMed, and OpenAlex. BibTeX export and shareable findings. Built on the IsonAI platform.

slate.isonai.net →
Forge
Coding agent

A coding agent for software development tasks. Available as a terminal CLI and as a browser interface with built-in sandbox execution.

www.isonai.net/forge →
Iris & Lumen
Image and video

Iris generates images and Lumen generates video. Both run on Epithre-operated GPUs in Jakarta and are available through IsonAI chat. Iris is additionally exposed on the developer API.

Inside www.isonai.net →
Platform
Developer API

An OpenAI-compatible API hosted in Jakarta. Endpoints for chat, embeddings, rerank, batch processing, retrieval, and webhooks. Billed in USD with data residency in Indonesia.

platform.epithre.com →
Models

What we serve.

Omni
Multimodal chat with vision and web search. The flagship general-purpose model.
PRME
Long-context reasoning for document workflows and large-corpus analysis.
LYT
Lightweight assistant for fast interactions.
Iris
Image generation with multi-reference editing.
Lumen
Text-to-video generation.
Embed
Multimodal text and image embedding for retrieval pipelines.
Rerank
Cross-encoder reranking for retrieval pipelines.

All models run on Epithre-operated GPUs in Jakarta. Pricing and account signup at platform.epithre.com. API references at docs.epithre.com.

Stack

From hardware to user.

Layer 4 · Top Applications
User-facing products built on the platform below. IsonAI, Slate, Forge, Iris, Lumen.
Layer 3 Developer API
OpenAI-compatible HTTP surface at api.epithre.com. Chat, embeddings, rerank, retrieval, batch, webhooks. USD billing.
Layer 2 Inference runtime
In-house quantized model serving with retrieval and embedding pipelines. Routes to Omni, PRME, LYT, Iris, Lumen, Embed, Rerank.
Layer 1 · Base Owned infrastructure
Physical GPU servers, storage, and networking in Jakarta, Indonesia. No third-party cloud in the production path.
Request POST /v1/chat/completions
$curl https://api.epithre.com/v1/chat/completions \
    -H "Authorization: Bearer $EPITHRE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "omni",
      "messages": [
        {"role": "user", "content": "Cara hitung PPh 21 buat freelancer?"}
      ]
    }'
Company

About us.

JAKARTA
Primary site Jakarta, Indonesia
Coordinates -6.21° S, 106.85° E
Site model Owned hardware

Epithre was founded in 2025 to build AI infrastructure for the Indonesian market. The company operates a fully integrated stack: physical server hardware, model serving runtimes, retrieval and embedding systems, and the consumer and developer applications that sit on top.

Engineering is performed entirely in-house. The team manages infrastructure provisioning, GPU operations, model quantization, inference runtime deployment, application engineering, and customer support. Core production systems do not run on third-party cloud providers.

Customer data is processed exclusively on Indonesian infrastructure operated by Epithre. The platform serves organizations in regulated sectors, including hospitality, public sector, and healthcare operations, where local data residency is a requirement.

Founded
2025
Headquarters
Jakarta, Indonesia
Data residency
Indonesia only
API surface
OpenAI-compatible
Languages
Indonesian, English
Sectors served
Hospitality, government, healthcare, software, research, creative
Contact

Get in touch.