Quick Start
Make your first API call in minutes. Get your API key, run inference, and fine-tune your first model.
API Reference
Full reference for every Pioneer endpoint — inference, training, datasets, evaluations, and projects.
Models
Browse available encoder (GLiNER) and decoder (LLM) models for fine-tuning and serverless inference.
Guides
Step-by-step walkthroughs for NER fine-tuning, LLM training, synthetic data generation, and more.
How Pioneer works
1
Choose a base model
Select from GLiNER encoder models for NER and extraction, or supported decoder families including Nemotron, DeepSeek V4 Flash, GLM 5.2, Claude 5, and GPT-5.
2
Upload or generate training data
Upload your labeled dataset or use Pioneer’s synthetic data generation to create training examples from scratch.
3
Start a training job
Submit a fine-tuning job via the API. Pioneer handles LoRA or full fine-tuning and reports F1, precision, and recall on completion.
4
Run inference
Call
POST /inference with your training job ID to serve predictions. Your fine-tuned model is deployed on-demand — no cold-start setup required.Pioneer also supports Adaptive Inference — a continuous improvement loop that automatically evaluates, retrains, and promotes model checkpoints based on live production traffic. See Adaptive Inference to learn more.
Key capabilities
NER Fine-tuning
Train GLiNER models on your entity types for extraction, classification, and structured JSON output.
LLM Fine-tuning
Fine-tune Nemotron 3.5 Lightning models on your domain data.
Synthetic Data
Generate labeled training data for NER and classification tasks — no manual annotation required.
Agent Skills
Give your AI coding agent (Cursor, Claude Code) full Pioneer API knowledge with a single SKILL.md file.