Deploytual
Deploy intelligence. Any data. Anywhere.
Problem
Most teams still bounce between spreadsheets, ad hoc SQL, notebook models, and slide decks. Answering a simple business question often means waiting on an engineer, copying data by hand, and hoping the last report is still accurate. Non-technical stakeholders rarely get explainable answers in one place.
Deploytual was built to close that gap: one deployable engine that connects to data, answers questions in plain English, runs AutoML, cleans data, and produces board-ready reports, without forcing every user through a notebook workflow.
Constraints
- Must run as a portable package (Docker / Kubernetes), not a locked SaaS-only tool.
- Answers must be explainable (show the SQL or Pandas code behind each result).
- Support mixed sources: CSV/Excel first, then SQL databases and APIs.
- Ship the smallest usable product path: upload → query → report → then harden auth, pipelines, and K8s.
Architecture
Ingestion, AI query, ML services, cleaning, reporting, and auth sit behind a FastAPI backend. The frontend is a single-page app. Pipelines chain ETL → ML → report, with WebSocket notifications on completion.
What I built
Backend and AI services
- FastAPI routes for upload, profiling, NL query, database chat, anomaly detection, Prophet forecasting, cleaning suggestions, PDF reports, auth, and pipeline execution.
- Natural-language layer that converts questions into SQL or Pandas, with the generated code shown for explainability.
- AutoML: Isolation Forest anomalies, Prophet forecasts, clustering; AI cleaning studio for missing values and outliers.
- JWT + Google OAuth; WebSocket toasts for pipeline and anomaly events.
Frontend and delivery
- Single-page frontend with Chart.js, storyteller (text-to-speech), and visual/NL pipeline builder.
- Docker Compose, GitHub Actions CI, production Helm chart, and deploy targets on Netlify (UI) and Render (API).
Results
Deploytual is a working product path for Forward Deployed and Data Engineering use cases: drop it on a client environment, connect data, answer questions with code transparency, and export executive PDFs without a custom BI project each time.
What I would do differently
- Introduce multi-user workspaces and RBAC earlier so auth models the real multi-tenant case.
- Add contract tests on profile and query schemas so frontend and backend stay aligned under rapid feature growth.
- Prioritize first-class PostgreSQL and MongoDB connectors before broader "any API" abstractions.
- Ship a short recorded NL-query demo clip for portfolio visitors who will not run the live app.
Open to internships and full-time roles
Actively looking for Data Engineer, Forward Deployed Engineer, and Analytics Engineer opportunities. If you are hiring, I would love to hear from you.