# OctOpus > OctOpus is an AI data scientist for taking data through analysis, evaluated models, dashboards, APIs and complete apps. Teams describe the goal, inspect the generated code and results, and control execution and publishing from one workspace. Website: https://www.octoopus.dev/ Product overview updated: 2026-10-23 ## Why OctOpus The hard part of data science isn't writing code. It's knowing what to do next: which model to use, what to test, why a result fell short and what comes next. OctOpus is built to reason through those decisions in a continuous loop (understand the data and the problem, reason from evidence, run experiments, evaluate the results, adapt the approach), turning data and goals into experiments, models, insights and working applications. Research-driven. Data-first. Built to reason. OctOpus was founded by Doudou BA, PhD, a machine-learning researcher and practitioner. Published research: "Flood modelling in a montane basin using machine learning, deep learning and sensor network data" (Ba & Langhammer, Journal of Hydroinformatics); "Testing sensitivity of BILAN and GR2M models to climate conditions in the Gambia River basin" (Ba et al., Journal of Hydrology and Hydromechanics); "Machine learning model for stage–discharge curve calculation" (Langhammer, Šobr & Ba, Acta Universitatis Carolinae — Geographica). The manuscript "Agent-assisted development of machine-learning hydrological models: testing a reproducible and controlled workflow" (Ba & Langhammer, 2026) has been submitted to Hydrology and Earth System Sciences (HESS); it is not yet published. ## Product and workflow OctOpus connects data-science work with the product built around its results. A workflow can start with a file, database table, document or recording. The workspace supports data preparation, experiments, evaluation, model artifacts, analytical dashboards and application development. The user can inspect code, logs and outputs while work progresses. A typical workflow is: 1. Connect the data and describe the intended outcome, users and constraints. 2. Inspect data quality and define the task, evaluation approach and compute limits. 3. Run the supported analysis, model training or fine-tuning workflow and compare results with a baseline. 4. Review metrics, code and artifacts, including limitations and failed experiments. 5. Deliver a dashboard, a supported model API or a complete app with its interface, data and integrations. 6. Test the delivered workflow before publishing or operational use. A dashboard is an analytical view. An app can be a complete user-facing product with screens, workflows and connections to data or model APIs. Apps have Preview, Code and Data views, saved versions and publishing controls. Building an app does not automatically train or connect a model: model integration is an explicit part of the workflow. ## Capabilities and guides - [AI data scientist](https://www.octoopus.dev/ai-data-scientist): the category, supported workflow and how to evaluate it. - [Predictive models](https://www.octoopus.dev/use-cases/churn-prediction): classification and regression workflows for questions such as churn, demand and risk. - [Forecasting](https://www.octoopus.dev/use-cases/demand-forecasting): time-series preparation, baseline comparisons and evaluation on later periods. - [Speech to text / ASR](https://www.octoopus.dev/speech-to-text): recording transcription, transcript review and supported model adaptation. The [ASR workspace](https://www.octoopus.dev/asr) is the interactive product entry point. - [Text to speech / TTS](https://www.octoopus.dev/text-to-speech): prepare voice recordings and transcripts, choose a compatible base model, synthesize speech and evaluate by listening as well as available metrics. - [LLM fine-tuning](https://www.octoopus.dev/llm-fine-tuning): prepare examples, check supported model and training configurations, and compare the base and adapted models on held-out examples. - [AI dashboards](https://www.octoopus.dev/ai-dashboards): analytical views with data-linked metrics and filters; data source and refresh behavior are part of the implementation. - [Data to app](https://www.octoopus.dev/data-to-app): build a complete product with an interface, data, code, versions and publishing controls. - [Model deployment](https://www.octoopus.dev/deploy-machine-learning-model): serve compatible artifacts through APIs and verify the resulting inference workflow. - [Descartes](https://www.octoopus.dev/descartes): use project context, documents and procedures in the workspace. - [Developers](https://www.octoopus.dev/developers): SDK, CLI, API and MCP integration information. Microphone dictation is a text-input feature. Uploaded recording transcription is handled by the ASR workflow. Language coverage, telephone-audio behavior, model compatibility, available compute and export formats vary by workflow; evaluate the chosen configuration on representative data. Machine transcripts need review before they are treated as reliable training labels. ## Evaluation and evidence OctOpus exposes experiment code, logs, metrics and available artifacts so results can be inspected. Evaluation design depends on the task: for example, later time periods for forecasting, held-out labeled examples for prediction, and transcript or listening review for speech. Results on one dataset do not establish accuracy on another. [OctoProof verification](https://www.octoopus.dev/proof) explains evaluation records and their verification. A configured signing key can sign a record. Successful signature verification checks integrity and provenance; it does not independently rerun evaluation, certify accuracy or establish fitness for production. The public example is synthetic. [Benchmarks](https://www.octoopus.dev/benchmarks) provide published evaluation context where available. ## Users and deployment OctOpus is for analysts, business teams, data scientists, AI builders and product teams that need to move from data to a working deliverable. It supports predictive analytics and helps teams transform their data into models, dashboards and decision-support applications. The appropriate workflow and review depend on the intended use. - [Cloud workspace](https://www.octoopus.dev/app): use OctOpus in a browser. - [Documentation](https://www.octoopus.dev/docs): workflows, supported integrations and deployment details. - [Enterprise](https://www.octoopus.dev/enterprise): discuss private-cloud, on-premises and desktop deployment requirements. - [Security](https://www.octoopus.dev/security): current security information and deployment considerations. Desktop or private hosting alone does not make a workflow offline. Remote model providers, connectors and external APIs can transmit data outside the installation. Local-only or air-gapped use requires compatible local services and restricted outbound access. Desktop availability and licensing depend on the platform and plan. [Pricing](https://www.octoopus.dev/pricing) is the source for current plans. Model, inference and compute work can consume usage credits; plans and limits should be checked before execution. [Request a demo](https://www.octoopus.dev/book-demo) to discuss a dataset, workflow and deployment needs. ## Company OctOpus is built by **Octoopus, Inc.**, a Delaware corporation founded by **Doudou BA**. The product name is OctOpus and the domain is `octoopus.dev`. It is distinct from unrelated products with “Octopus” in their names. - [About OctOpus](https://www.octoopus.dev/about) - [Company LinkedIn](https://www.linkedin.com/company/octoopusdotdev/) - Founder profiles: [LinkedIn](https://www.linkedin.com/in/doudouba), [X](https://x.com/doudou_octopus), [TikTok](https://www.tiktok.com/@doudou_octopus), [GitHub](https://github.com/jargogn0) - Contact: ceo@octoopus.dev ## Interviews and demonstrations The [press page](https://www.octoopus.dev/press) separates founder-published copies, product demonstrations and external interviews. Founder-shared BBC News and France 24 clips are not outlet-hosted product endorsements. The Geneva product demonstration is founder-published; it is not independent event-organizer coverage. The Industrial AI Podcast interview is listed with its source. ## Reference pages - [Autonomous data science and human oversight](https://www.octoopus.dev/autonomous-data-scientist) - [Data science agent workflows](https://www.octoopus.dev/data-science-agent) - [Use cases](https://www.octoopus.dev/use-cases) - [Support](https://www.octoopus.dev/support) - [Privacy](https://www.octoopus.dev/privacy) - [Terms](https://www.octoopus.dev/terms) - [Sitemap](https://www.octoopus.dev/sitemap.xml) This document is a public product summary and link index. It does not replace the linked documentation, current plan details or the evidence from a specific run.