Foundation models are expensive.
A trading agent that calls GPT-4 for every price prediction burns budget fast.
Lightweight forecasters are cheap to run but expensive to train, especially when you only have a handful of examples.
MetaCaster introduces a meta-harness architecture where agents don't forecast directly.
Instead, they train specialized lightweight models on-demand from few-shot examples and textual context.
This is not another AutoML wrapper.
The meta-agent orchestrates data generation, architecture selection, and training loops to produce task-specific forecasters in minutes.
The result is a deployable model that runs inference without touching the foundation layer again.
The Economic Gap Time-series forecasting in production faces a resource trap: Foundation models (TimeGPT, Chronos) deliver strong zero-shot performance but cost $0.002 to $0.02 per prediction at scale.
Lightweight forecasters (PatchTST, DLinear, FEDformer) run for pennies but need thousands of training samples and hours of GPU time.
Few-shot scenarios (new trading pairs, emerging markets, privacy-sensitive health data) don't have enough history to train from scratch.
MetaCaster targets the intersection: resource-constrained environments where you need specialized models but can't afford foundation API calls or long training cycles.
Meta-Harness Architecture The system has three layers:
1.
Meta-Agent Orchestrator The top-level agent receives a few-shot time series (as few as 5-10 examples) and optional textual context (domain descriptions, seasonality hints).
It decides: Which lightweight forecaster architecture to instantiate (PatchTST, DLinear, Autoformer, etc.) What synthetic data generation strategy to apply How to configure the training harness (learning rate, epochs, augmentation) The meta-agent uses a learned policy, not heuristics.
It's pre-trained on a meta-dataset of diverse forecasting tasks so it generalizes to new domains.
2.
Data Generation Agents These agents expand the few-shot examples into a trainable dataset.
Strategies include: Perturbation agents: Add noise, shift phases, scale amplitudes while preserving statistical properties.
Interpolation agents: Generate intermediate sequences between observed samples.
Context-guided synthesis: Use textual hints (e.g., "weekly retail sales with holiday spikes") to steer generation.
The generated data is not generic.
It's tuned to the target task's distribution based on the meta-agent's analysis of the few-shot examples.
3.
Lightweight Forecaster Training Loop The meta-harness spawns a training job with the selected architecture and synthetic dataset.
This is a standard supervised loop, but the harness monitors: Validation loss on held-out few-shot examples: Prevents overfitting to synthetic data.
Training time budget: Stops early if the model converges or hits a wall-clock limit.
Architecture-specific hyperparameters: Each forecaster family (transformer, MLP, CNN) has different sensitivities.
Once trained, the lightweight model is serialized and cached.
The meta-agent never touches it again unless the task distribution shifts.
Training Loop Boundary The critical design choice is where the meta-agent stops and the forecaster starts.
MetaCaster uses a clean separation: Meta-agent: Operates in the space of architectures, data strategies, and hyperparameters.
It does not see raw time-series values during meta-training.
Forecaster: Operates in the space of time-series predictions.
It does not know it was trained by an agent.
This boundary matters for versioning and reproducibility.
You can snapshot the trained forecaster and deploy it independently.
The meta-agent is only needed when you want to train a new model or retrain an existing one.
Versioning and Caching MetaCaster includes a model registry that hashes: The few-shot input examples The textual context The selected architecture and hyperparameters If an agent requests a forecaster for a task it's seen before, the system returns the cach