Skip to main content
Pioneer is an AI fine-tuning and inference platform built for teams that want more control over their models without managing infrastructure. You bring your data; Pioneer handles training, evaluation, and serving. Connect through the Pioneer REST API or use drop-in OpenAI- and Anthropic-compatible endpoints to fit into any existing stack.

​
What you can do with Pioneer

NER fine-tuning

Train GLiNER encoder models on your own entity types. Extract named entities, classify text, and produce structured JSON output from unstructured text.

LLM fine-tuning

Fine-tune decoder models — Qwen, Llama, DeepSeek, and others — on your domain data using LoRA. Download the weights when training is complete.

Inference

Run predictions against any base model or your own fine-tuned model via POST /inference. Pioneer also exposes OpenAI- and Anthropic-compatible endpoints for drop-in compatibility.

Synthetic data generation

Generate labeled training examples for NER and classification tasks without manual annotation. Describe your domain and Pioneer produces ready-to-use datasets.

​
Supported model families

Pioneer supports two classes of models: encoder models for structured extraction tasks, and decoder models for generative tasks. Encoder models
  • GLiNER — A small, efficient model purpose-built for named entity recognition, text classification, and structured JSON extraction. GLiNER is the recommended starting point for agent text processing, document parsing, and routing workflows.
Decoder models (LLMs)
  • Qwen — Strong at coding, multilingual tasks, and complex multi-step reasoning. Ideal for global products.
  • Llama — Meta’s open-source model family. Well-suited for RAG, summarization, and general-purpose chat.
  • DeepSeek — Capable at code generation, structured reasoning, and agentic planning tasks.
To see all available base models, call GET /base-models. You can filter by task type or inference support.
Pioneer supports Adaptive Inference on the Research plan — a continuous improvement loop that automatically evaluates live traffic, generates training data, and promotes improved checkpoints. See Adaptive Inference to learn more.

​
How it fits into your workflow

Pioneer follows a straightforward lifecycle: upload or generate data → start a training job → evaluate the result → run inference with your trained model.
1

Choose a base model

Pick a GLiNER encoder for extraction and classification, or a decoder LLM for generative tasks. Use GET /base-models to browse what’s available.
2

Prepare your data

Upload a labeled dataset or use Pioneer’s synthetic data generation to create training examples from a domain description and label list.
3

Fine-tune

Start a training job with POST /felix/training-jobs. Pioneer runs LoRA fine-tuning and returns F1, precision, and recall metrics on completion.
4

Evaluate and serve

Run an evaluation with POST /felix/evaluations to benchmark your model, then serve predictions with POST /inference using your training job ID.

​
Next steps

Quickstart

Make your first API call in under five minutes.

Authentication

Generate your API key and authenticate requests.