Neotropy Forge · coming soon

No-code fine-tuning. API-first everything.

Forge trains open-source models on your proprietary data — a visual pipeline for humans, a clean REST surface for your infrastructure, from first upload to exported weights.

Get Notified at Launch

supportsLlamaQwenMistralGemma+ any HF model

01 — Estimator

Know the price before you press train

Pick a model, a method, and a dataset. The estimator prices the run from public throughput numbers — the exact quote appears before you commit.

configuration

base model

method

dataset size

40,000 rows

~600 tokens/row · 2048 ctx · 3 epochs

estimate

32M

trainable params

11 GB

peak VRAM

2.2 h

est. wall time

$8.56

est. compute cost

recommended hardware: 1× RTX 4090 · 24.0M tokens

Ballpark figures from public throughput numbers. Forge quotes exact price-per-run before you spend a cent.

02 — Capabilities

Everything between corpus and checkpoint

Full & QLoRA support

Full-parameter runs when you own the GPUs, rank-stabilized QLoRA when you don't. Same lineage, same evals, one click apart.

FP16NF4 4-bitr-stabilized

Automated synthetic data cleaning

Dedup, de-PII, and difficulty-scoring pipelines filter synthetic and raw corpora before a single training step runs.

dedupde-PIIcurricula

Evaluation benchmarks vs base

Every run is scored against its base model on your benchmarks and public suites — with regression gates before export.

vs baseregression gates

1-click model export

Ship merged or adapter-only weights wherever they need to live: GGUF for the edge, vLLM for serving, or your HF org.

GGUFvLLMHuggingFace

03 — Two front doors

Click it or script it — same pipeline

no-code

Upload corpus412 MB parsed

Auto-clean22% noise removed

Train QLoRAstep 380/480 · loss 0.81

Exportqueued · gguf + vllm

api-first

$ curl -X POST https://api.neotropy.ai/v1/fine-tunes \

-H "Authorization: Bearer $NT_KEY" \

-d { "base": "qwen2.5-7b", "method": "qlora",

"dataset": "ds_8841", "export": ["gguf","vllm"] }

→ run ft_2214 · queued · est $5.40

Every button in the UI is a documented endpoint. Webhooks, SDKs, and OpenAPI spec ship at launch.

coming soon

AI Forge — not live yet — but it is close.

We are finishing hardening, safety review, and onboarding tooling. Leave your email and we will notify you the moment it opens — one email, at launch, nothing else.

SOC 2 Type II on-prem available data never leaves your VPC