Neo
Back to use cases

Workflow

Train & Fine-Tune Models

LoRA adapters, supervised fine-tuning, knowledge distillation, RLHF, and quantization, from raw data to deployment-ready weights.

The problem

Fine-tuning and adaptation work is repetitive: preparing data, picking hyperparameters, running training, and validating weights, all while hardware and cost constraints keep shifting.

  • A training run finishes and nobody's sure the hyperparameters were actually the right ones.
  • Compressing a model for edge deployment usually means guessing how much accuracy you can afford to lose.
  • Nobody's tracking what a training run actually costs, in dollars or in carbon.

How NEO helps

NEO drives the loop end to end: curated datasets, adapter and full fine-tune runs, distillation and quantization checks, and reproducible artifacts you can promote to staging without babysitting notebooks.

  • Runs LoRA adapters end-to-end and quantizes checkpoints to GGUF for offline deployment
  • Schedules training around real-time grid carbon intensity and tracks emissions automatically
  • Applies quantization-aware training so compression trades away the least accuracy possible

Scenarios (3)

9x Model Compression for Edge
9x model compression

Train & Fine-Tune Models

9x Model Compression for Edge

INT8 quantization-aware training compressing MobileNetV2 from 23.5MB to 2.6MB with only 3.8% accuracy drop.

Read case study
Medical Reasoning via LoRA Fine-Tuning
7.8GB offline GGUF

Train & Fine-Tune Models

Medical Reasoning via LoRA Fine-Tuning

Gemma-3-12B adapted for symptom triage, drug interactions, and lab interpretation: 7.8GB GGUF for fully offline clinical use.

Read case study
Carbon-Aware Training Scheduler
43% CO2 reduction

Train & Fine-Tune Models

Carbon-Aware Training Scheduler

Schedules training around grid carbon intensity using CodeCarbon tracking: 43% CO2 reduction with accuracy within 0.3% of baseline.

Read case study

Run this workflow in your workspace

Install NEO in VS Code or Cursor and ship training, evals, and deployment from chat.

Get started
CursorCursor
Claude CodeClaude Code
CodexCodex
WindsurfWindsurf
ZedZed
AntigravityAntigravity
ContinueContinue
CursorCursor
Claude CodeClaude Code
CodexCodex
WindsurfWindsurf
ZedZed
AntigravityAntigravity
ContinueContinue
eval_pipeline.py
  def run_eval(model, tasks):-     results = model.predict(tasks)+     results = batch_eval(model, tasks)+     log_regression(results)      return score(results)