HerbDev model-development proof of concept

Multilayer Perceptron (MLP)

This case study shows the complete model-training workflow: establish a baseline, inspect evidence, make a controlled optimization, rerun the model, and preserve the resulting artifacts.

Baseline

validation_accuracy	0.65
test_accuracy	0.5875
train_seconds	0.1277

Optimization decision

Increase representation capacity and train longer at a lower learning rate after the baseline underfit the nonlinear boundary.

Optimized run

validation_accuracy	0.95
test_accuracy	0.9625
train_seconds	0.49875

Training evidence

Baseline and optimized training loss comparison

Execution environment

platform	Linux-6.18.35-x86_64-with-glibc2.41
machine	x86_64
python	3.13.5
torch	2.10.0+cpu
device	cpu
mps_available	False
cuda_available	False
seed	42

What this demonstrates for a client

Model development is not a single training command. The useful work is creating a reproducible dataset, choosing a defensible baseline, measuring behavior, changing one thing for a reason, retesting, and preserving enough evidence to explain what happened. The same process scales to client data, larger models, and production infrastructure.

Have a model or dataset that needs improvement?

The next step is to define the business outcome, available data, constraints, and success metric, then design the smallest experiment that can prove whether machine learning is worth pursuing.