ML / Software · Applied ML project
Financial Decision System
Structured risk inference with explanation layer
Built a financial-risk prototype that prepares structured inputs, runs a prediction model, and then explains the result in plain language. The explanation layer is separate from the model so generated text cannot replace or change the underlying prediction.
Automated data preparation and model inference instead of relying on manual notebook steps.
Kept the model output separate from the LLM-generated explanation.
Used both C++ and Python across the prototype.
Input transformations had to stay consistent every time the model ran.
The explanation needed to communicate the result without making the language model appear to be the prediction model itself.
Built an end-to-end prototype from structured financial inputs to prediction and user-facing explanation.