Early access
Replace an LLM API callwith a model you own.
For one repetitive task, like routing tickets or extracting fields, SynqTune trains a small model on your own examples and tests it against the AI you use today before you switch. Your data and weights never leave your AWS account.
Or start free: look around with sample runs first, no AWS account needed.
$claude mcp add synqtune -- uvx synqtune-mcpfor coding agentsWhat you see before switching
Accuracy on your test examples
higher is better
Latency per request
lower is better
Cost per 1,000 requests
lower is better
Verdict after every example is scored
Ready to replace, close, or not ready.
Ready to replace
Within your tolerance on every label, at a fraction of the cost.
Illustration. Your report shows your own numbers, measured in your account.
- Platform fee
- $0 in early access
- Runs in
- Your AWS account
- Data & weights
- Stay in your S3
- Switch when
- The report says ready
Design partners
We'll run your first comparison with you
Early access is hands-on. Bring one task you pay an LLM API for; we set it up together, in your AWS account, and you see the report on your own data before deciding anything.
A 30-minute call
Tell us the task you'd move off the API. We set up the run with you, on the call.
Run in your AWS account
Your examples never leave it. You get the comparison report on your own data, free.
You shape what we build
Early partners' requests go to the top of the roadmap, and launch pricing is locked in.
Prefer email? support@synqtune.com
[ Built on ]
- Amazon EC2
- Amazon S3
- AWS CloudFormation
- AWS KMS
- Unsloth
- Hugging Face PEFT
- llama.cpp
In 50 words
What SynqTune does
In plain words
Companies pay AI providers for millions of small, repetitive tasks, like sorting support tickets. SynqTune trains a small AI model on your own examples, inside your AWS account, then tests it against the AI you pay for today. If it matches, you can run your own cheaper, private model instead.
For engineers
SynqTune fine-tunes small open models with LoRA on your labeled data, on GPUs in your own AWS account, starting from task templates. It then runs a parity report against your current LLM on held-out examples and gives a clear verdict. Your data and weights stay in your S3. No ML team needed.
Think of it like
Replacing an expensive consultant you call a million times a month with an in-house specialist trained for that one job, after testing that the specialist is just as good.
Six steps
Peel back a run, one step at a time
From picking a task to running a model you own. Everything after step 1 happens in your AWS account.
[ 06 ] Switch
Download the adapter or GGUF from your bucket and run it where you choose.
[ 05 ] Compare
Both models answer the same test examples. The report says whether yours is ready.
[ 04 ] Fine-tune
A small open model trains on a GPU in your AWS account. Checkpoints save to your S3.
[ 03 ] Clean and check
Remove duplicates, bad answers and personal data. Try each step on a few records first.
[ 02 ] Bring examples
A file in S3, an upload, or the prompts you already send to an LLM, from logs or Langfuse.
[ 01 ] Pick a task
Classification, extraction, tagging, redaction or summarization. Fill in your labels or schema.
- 01Pick a taskClassification, extraction, tagging, redaction or summarization. Fill in your labels or schema.
- 02Bring examplesA file in S3, an upload, or the prompts you already send to an LLM, from logs or Langfuse.
- 03Clean and checkRemove duplicates, bad answers and personal data. Try each step on a few records first.
- 04Fine-tuneA small open model trains on a GPU in your AWS account. Checkpoints save to your S3.
- 05CompareBoth models answer the same test examples. The report says whether yours is ready.
- 06SwitchDownload the adapter or GGUF from your bucket and run it where you choose.
Use cases
Pick a task, get a pipeline
Each task is a template in the builder: fill in your labels or schema, point it at your examples, and it lays out the steps for you.
Classification and routing
Ticket routing, intent detection, triage
You fill in
- Labelsbilling, refund, shipping, technical
Your examples
Each row: the input text (instruction or prompt column) and its label (output or response column).
The pipeline it builds
- Your examples
- Remove duplicates
- Fine-tune
- Evaluate vs your LLM
- Save model
Your examples can be a file in S3, an upload, an LLM logs file or a Langfuse import. Add or remove steps on the canvas before you launch.
Checked against your current LLM
Accuracy and macro-F1 on held-out examples, for the fine-tuned model and the LLM you use today
In a regulated industry? Every task runs in your own AWS account.How data stays in your account
Security and data residency
SynqTune coordinates the work. Your account does it.
Your examples, checkpoints and model weights are read and written in your AWS account. SynqTune starts and stops the instances through one IAM role you can delete at any time, and every permission it has is listed publicly.
Compute in your account
EC2 instances you own, terminated when the run ends or sits idle.
Files in your bucket
Datasets, checkpoints and weights are written only to your S3.
One scoped role, revocable
Only its own bucket and the instances it tags. It can't read your saved keys. Delete the role to revoke it.
SynqTune control plane
Pipeline settings · run status · where provider keys are kept
Your AWS account
S3 data
raw/ · data/
EC2 workers
CPU + GPU, spot
S3 model
model/<run>/
FAQ
Common questions
More on setup, every pipeline step and troubleshooting is in the documentation.
Why would a small model be as good as a big one?
Because it only does one narrow job. A frontier LLM is a generalist; a small model fine-tuned on thousands of your own examples becomes a specialist at your task, such as routing your tickets into your categories. It won't write essays, but on that one task it can match the big model, and the comparison report shows whether it does.
Can I use my model in production today?
Training and testing are ready today, free during early access. You can download the weights (a LoRA adapter, or a GGUF file for Ollama or llama.cpp) and run them yourself, or chat with the model on an instance in your account. A hosted endpoint your applications call directly is next on the roadmap, and early-access teams help shape it.
How do I know the small model is good enough to switch?
Add a comparison to the Evaluate step: your current LLM, with your production prompt, answers the same held-out examples as the fine-tuned model. The report scores both for your task, shows where the small model is weaker, and ends with ready to replace, close, or not ready.
Why fine-tune a small model instead of using RAG or a frontier LLM?
They solve different problems. A frontier LLM is best for open-ended work, and RAG for answering from documents that change. For a narrow, repeated task with examples of good answers, a fine-tuned small model learns the task itself, so each request carries only the input, costs GPU time instead of per-token fees, and stays in your AWS account. They also combine: retrieval can feed facts to your fine-tuned model.
Does my data leave my AWS account?
Training data and weights stay in your S3, and so do the evaluation examples. SynqTune stores pipeline settings, run status and evaluation scores and counts. If you turn on LLM grading or the comparison with your current LLM, records are sent to the provider you choose.
Do I need ML or DevOps experience?
No. Setup is one CloudFormation stack launched from the AWS console, and pipelines start from task templates with defaults. Every setting is visible and explained before you launch.
Free early access
Train the model your product needs
Connect your AWS account in about five minutes, pick a task, and see how a model trained on your examples compares with the LLM you use today.