MyQuery | AI-Powered Analytics & Dashboard Generation Platform

MyQuery is an AI-powered business analytics platform that lets non-technical users connect their databases and ask questions in natural language — no SQL required.

A business analyst can type "Show me monthly sales by region for the last 12 months" in any language and the platform handles everything: generates the SQL, executes the query, builds the visualization, assembles a dashboard, and surfaces a written summary of the insights. The whole pipeline runs automatically.

I worked across the full platform lifecycle — backend, infrastructure, AI layer, and architecture.

Built the database connectivity layer supporting 20+ SQL and NoSQL data sources including PostgreSQL, MySQL, Oracle, MongoDB, Snowflake, BigQuery, ClickHouse, Elasticsearch, Salesforce, Odoo, and Google Sheets. Each integration required handling different SQL dialects, metadata systems, connection methods, and authentication mechanisms while presenting a unified experience to the AI layer above it.

Also built the query generation pipelines, dashboard generation services, authentication and authorization features, and self-hosted licensing infrastructure.

Integrated OpenAI, Anthropic, Gemini, and Groq as interchangeable model backends. Built the prompt engineering layer, tool calling workflows, context management system, and dashboard generation pipelines on top of those integrations.

The most impactful work here was redesigning how schema context gets passed to LLMs. Originally the system sent full schemas — large tables, excessive metadata — which drove token usage to ~20k input / ~10k output tokens per query. I rebuilt the context selection strategy to intelligently surface only the schema information relevant to each query.

Result: 80%+ reduction in token consumption. Input dropped from ~20k to ~3k tokens, output from ~10k to ~1k. Costs dropped significantly and response quality actually improved because the model had less noise to work through.

Designed and owned the full deployment stack:

Dockerized deployments and Docker Compose environments

NGINX reverse proxies with SSL automation

GitHub Actions CI/CD pipelines

Blue-green deployments with automatic rollback

Health checks and monitoring

VPS and cloud infrastructure across Contabo, Google Cloud Run, and Railway

Also built a self-hosted enterprise edition deployable on Windows, Linux, and macOS via Docker — including licensing, activation workflows, update management, and cross-platform compatibility.

Owned architecture decisions, technical reviews, PR reviews, production troubleshooting, and deployment operations across multiple product iterations as part of a startup engineering team.

20+ data source integrations across SQL, NoSQL, and SaaS systems

80%+ reduction in AI token usage through context compression (20k → 3k input tokens)

Multilingual natural language support — users can query in any language

Self-hosted enterprise edition deployable on any major OS via Docker

Multi-cloud deployments across Contabo VPS, Google Cloud Run, and Railway