Theovya Prediction Engine

A trained model from a prompt, with honest metrics attached.

Builds, trains, and validates real machine-learning models from the user’s own data, guided entirely by prompts, and returns honest held-out metrics with every model.

Status

in progress

Answers

what will happen

Builds on

Analytics Engine results

Metrics

held-out, with every model

How it works

Prompt in, trained model out.

Forecast demand per SKU for the next eight weeks.

  1. 01Three inputs: results already surfaced by the Analytics Engine, the underlying database, and the user’s prompt.
  2. 02Our proprietary agentic structure reads the shape of the data (task type, target, features, volume) and composes a prediction specification.
  3. 03It selects candidate models, or an ensemble, from a curated catalog, then runs a real ML workflow: preparation, training, validation, held-out testing.
  4. 04The lifecycle stays prompt-driven: promote a model into production, wire it into a dashboard, schedule re-forecasts and alerts.

Why the forecast is measured, not asserted

The language model never generates a prediction. Our agentic structure generates a plan: which models to try, how to prepare the data, how to split train, validation, and test.

The predictions themselves are computed by statistical and machine-learning models fitted to real company data, and the engine reports genuine held-out metrics: measured accuracy, not the model’s confidence in its own prose.

Validated by current research

AutoML-Agent (ICML 2025) shows multi-agent LLM systems taking a plain-language task description through to deployment-ready models across the full pipeline; DS-Agent shows LLM agents with case-based reasoning automating iterative data-science workflows. Theovya applies that architecture directly on top of a company’s live database.

The three engines

Three engines. Independent, composable, built to compound.

Each engine answers the next question on the maturity ladder. A business runs one, some, or all three, configured to what it actually needs, not a fixed bundle.

01 Available today

Answers what happened

Analytics Engine

Turns plain-language questions into live queries, charts, dashboards, and scheduled alerts, grounded in the company’s real schema, with the exact query behind every answer.

Explore the Analytics Engine
02 In progress

Answers what will happen

Prediction Engine

Builds, trains, and validates real machine-learning models from the user’s own data, guided entirely by prompts, and returns honest held-out metrics with every model.

You are here

03 On the roadmap

Answers what should we do

Optimization Engine

Translates business goals and constraints into formal optimization problems, solved by exact methods where optimality is provable and heuristics where scale demands it.

Explore the Optimization Engine

All three share one design principle: the language model never invents an answer. It only translates intent into verifiable computation.

Use cases

The Prediction Engine, applied.

Each case picks up from what happened, forecasts what comes next, and hands that forecast to optimization.

08 cases

AnalyticsPredictionOptimization

Most forecasts are never commissioned at all.

The Analytics Engine answers what happened. In every business in these cases, the next question is what will happen - and that is where the queue gets longest. A forecast has traditionally meant a project: a data scientist to frame it, an engineer to extract and prepare the data, weeks of model building, a validation report few people outside the team can read, and a model that goes stale soon after it is delivered.

The Prediction Engine turns the request into the project. Describe what you want to predict in plain language. It composes a specification you can read, trains candidate models on your own data, tests them on data they have never seen, and reports honest held-out accuracy with every forecast. The language model plans the work; trained models compute the numbers.

Start with one business unit.

The Analytics Engine is live and connects read-only to systems already in place. As the Prediction Engine ships, the same deployment expands to forecasting on the same footing.

Request a conversation

Theovya Prediction Engine

A trained model from a prompt, with honest metrics attached.

Builds, trains, and validates real machine-learning models from the user’s own data, guided entirely by prompts, and returns honest held-out metrics with every model.

REQUEST A CONVERSATION See how it works

Status

in progress

Answers

what will happen

Builds on

Analytics Engine results

Metrics

held-out, with every model

How it works

Prompt in, trained model out.

Forecast demand per SKU for the next eight weeks.

  1. 01Three inputs: results already surfaced by the Analytics Engine, the underlying database, and the user’s prompt.
  2. 02Our proprietary agentic structure reads the shape of the data (task type, target, features, volume) and composes a prediction specification.
  3. 03It selects candidate models, or an ensemble, from a curated catalog, then runs a real ML workflow: preparation, training, validation, held-out testing.
  4. 04The lifecycle stays prompt-driven: promote a model into production, wire it into a dashboard, schedule re-forecasts and alerts.

Why the forecast is measured, not asserted

The language model never generates a prediction. Our agentic structure generates a plan: which models to try, how to prepare the data, how to split train, validation, and test.

The predictions themselves are computed by statistical and machine-learning models fitted to real company data, and the engine reports genuine held-out metrics: measured accuracy, not the model’s confidence in its own prose.

Validated by current research

AutoML-Agent (ICML 2025) shows multi-agent LLM systems taking a plain-language task description through to deployment-ready models across the full pipeline; DS-Agent shows LLM agents with case-based reasoning automating iterative data-science workflows. Theovya applies that architecture directly on top of a company’s live database.

The three engines

Three engines. Independent, composable, built to compound.

Each engine answers the next question on the maturity ladder. A business runs one, some, or all three, configured to what it actually needs, not a fixed bundle.

01 Available today

Answers what happened

Analytics Engine

Turns plain-language questions into live queries, charts, dashboards, and scheduled alerts, grounded in the company’s real schema, with the exact query behind every answer.

Explore the Analytics Engine
02 In progress

Answers what will happen

Prediction Engine

Builds, trains, and validates real machine-learning models from the user’s own data, guided entirely by prompts, and returns honest held-out metrics with every model.

You are here

03 On the roadmap

Answers what should we do

Optimization Engine

Translates business goals and constraints into formal optimization problems, solved by exact methods where optimality is provable and heuristics where scale demands it.

Explore the Optimization Engine

All three share one design principle: the language model never invents an answer. It only translates intent into verifiable computation.

Use cases

The Prediction Engine, applied.

Each case picks up from what happened, forecasts what comes next, and hands that forecast to optimization.

08 cases

AnalyticsPredictionOptimization

Most forecasts are never commissioned at all.

The Analytics Engine answers what happened. In every business in these cases, the next question is what will happen - and that is where the queue gets longest. A forecast has traditionally meant a project: a data scientist to frame it, an engineer to extract and prepare the data, weeks of model building, a validation report few people outside the team can read, and a model that goes stale soon after it is delivered.

The Prediction Engine turns the request into the project. Describe what you want to predict in plain language. It composes a specification you can read, trains candidate models on your own data, tests them on data they have never seen, and reports honest held-out accuracy with every forecast. The language model plans the work; trained models compute the numbers.

Start with one business unit.

The Analytics Engine is live and connects read-only to systems already in place. As the Prediction Engine ships, the same deployment expands to forecasting on the same footing.

Request a conversation