Platform

Built for the full ML stack.

Inference API

One unified API for 200+ open and proprietary models. Streaming, function calling, JSON mode, and tool use built in.

OpenAI-compatible endpoints
Automatic batching & caching
Per-request observability

# curl example
curl https://api.nexus.ai/v1/chat \\
  -H "Authorization: Bearer $KEY" \\
  -d '{"model":"nexus-70b","prompt":"Hi"}'

Fine-tuning

Take any base model, run a LoRA or full fine-tune on your dataset, and deploy in minutes. No GPU provisioning required.

LoRA, QLoRA, and full fine-tunes
Bring-your-own-data via S3 or upload
One-command deploy from CLI

$ nexus tune create \\
  --base nexus-llama-8b \\
  --data ./train.jsonl \\
  --epochs 3

Observability

Trace every request, score every output, debug every prompt. Cost, latency, and quality metrics in one dashboard.

Prompt-level traces
Eval pipelines & A/B testing
PII redaction & audit logs

# SDK
client.trace(
  name="summarize",
  metadata={"user": user.id}
)