Bayesian optimization
Learn as you search. The optimizer uses completed trial results to choose promising parameter combinations for the next runs.
Find better configurations,without managing trials
Choose what to vary and what a good result looks like. HyperOptimizer runs your containers, learns from completed trials, and puts the results side by side.
Benefits
Spend your time on the code. Let us run the search.
Turn manual parameter tuning into an experiment you can follow, compare, and repeat. Bring a model, trading strategy, simulation, or another containerized workload.
Skip the cluster setup
Bring a Docker image and define your experiment. We schedule the trials, run the containers, and collect the results.
Your code, in containers
Keep your language, libraries, and dependencies. Package them in a Docker image and run a separate container for each trial.
Visual results dashboard
Follow the search as results arrive. Compare metrics, inspect trial logs, and use Pareto frontiers to explore trade-offs between objectives.
Parallel trial execution
Explore several configurations at once, with 2 trial slots on Free, 10 on Starter, and 50 on Pro. Each trial draws from your available compute.
Flexible parameter spaces
Integer, float, categorical: define any combination of hyperparameters and their ranges. The optimizer handles the sampling and search strategy automatically.
How it works
Three steps. No SDK.
Docker in, metrics out, ranked results on a dashboard. That's the whole contract.
1
Build a Docker image
Package your parameterized workload in a Docker container. Read trial arguments and emit objective metrics at runtime.
2
Parse args & print metrics
Read --hpo-* flags at runtime and emit metrics with the hpo.metrics. prefix.
3
View results in the dashboard
We run parallel trials, the optimizer suggests the next parameter set, and you see the live leaderboard as your experiment runs.
Strengths
Built for real workloads
No vendor lock-in, no proprietary SDK. Just Docker, stdout, and CLI args.
Docker-native
Use any language that can read CLI arguments and print metrics from a container. No SDK or client library is required.
A small integration surface
Read parameter values from CLI arguments and print metrics to stdout. Use the libraries you already know.
Multi-objective optimization
Optimize for multiple metrics simultaneously. Pareto frontiers help you balance competing objectives like return vs. drawdown.
Fault-tolerant trials
Inspect a failed trial without losing sight of the rest of the search. Completed results stay available within your plan’s history window.
Integrations
Works with your stack
HyperOptimizer is Docker-native. These guides show how specific frameworks integrate.
Docker
Bring a containerized workload that accepts parameters and emits metrics.
NautilusTrader
High-performance algorithmic trading optimization with Nautilus backtests.
Freqtrade
Optimize your crypto trading bot hyperparameters with managed infrastructure.
Soon
Bring your own cloud
Run trials in your cloud. Coming soon.
Ready to optimize?
Start with 200 compute minutes, free once per account. Need help connecting your framework? Let us know.