Defog - AI-powered text-to-SQL platform for enterprise data queries

What is Defog

Have you ever needed to answer a critical business question from your data, only to find yourself waiting in line for a data analyst? Or perhaps you've stared at a complex BI dashboard, clicking through dozens of filters just to get the insight you need. You're not alone. Across enterprises of every size, the gap between having data and actually understanding it has become a major bottleneck.

This is exactly the problem Defog was built to solve.

Defog is an AI-powered Text-to-SQL platform that lets you query your databases using plain English—no SQL knowledge required. Whether you're a marketing manager looking for campaign performance insights, a product owner checking user engagement metrics, or an executive needing quarterly numbers, Defog puts the power of data queries directly in your hands.

Under the hood, Defog is powered by SQLCoder, an open-source large language model specifically fine-tuned for converting natural language into precise SQL queries. This isn't a generic AI chatbot adapted for data work—SQLCoder was built from the ground up for one purpose: understanding your database schema and generating accurate, executable SQL.

The approach is resonating with the market. SQLCoder has been downloaded over 300,000 times, and industry leaders are taking notice. As Hugging Face CEO Clement Delangue noted, "Open models are now starting to surpass GPT-4 for specialized tasks. For example, the best open source model for SQL is SQLCoder-34B."

What makes Defog different from traditional BI tools? Instead of navigating complex dashboards and clicking through layers of filters, you simply ask your question in plain English. "What were our top-performing marketing channels last quarter?" Defog translates that into the exact SQL query needed and returns the insights you need—in seconds.

Core Features That Set Defog Apart

You need a solution that actually delivers results, not just impressive-sounding features. Here's what Defog brings to your data workflow.

SQLCoder: The Engine Behind the Magic

At the heart of Defog lies SQLCoder, a state-of-the-art Text-to-SQL large language model that has been fine-tuned specifically for database queries. Available in three parameter sizes—8b, 14b, and 32b—you can choose the right balance between speed and capability for your infrastructure. In multiple benchmark evaluations, SQLCoder has outperformed GPT-4 on SQL generation tasks, giving you more accurate results where it matters most.

You can use it to generate complex JOIN queries, subqueries, aggregations, and window functions—all from natural language descriptions of what you need.

Universal Data Source Connectivity

Your data likely lives across multiple systems—Snowflake for analytics, PostgreSQL for application data, SQL Server for legacy systems, MySQL for web applications. Defog connects to them all natively, without requiring custom integrations or data pipelines.

Most teams are up and running within 5 minutes of installation. Simply point Defog at your database, and it automatically understands your schema structure.

Deployment That Fits Your Workflow

How your team works should dictate how you deploy, not the other way around. Defog offers multiple deployment options: a desktop application for individual users, a Slack bot so teams can query directly from their communication hub, one-click deployment via AWS or GCP Marketplace, and Docker containers for custom infrastructure setups.

Whether you're a small team wanting quick adoption or an enterprise with strict infrastructure requirements, Defog adapts to you.

Privacy That Actually Means Privacy

This is where Defog stands apart from cloud-based AI query tools. Your data never leaves your environment. SQL queries execute directly within your infrastructure, and the self-hosted option ensures your most sensitive data stays behind your own firewall.

For industries with strict compliance requirements—financial services, healthcare, government—this architecture isn't a nice-to-have, it's essential. Defog supports SSO authentication integration and can comply with MSA and DPA requirements for enterprise contracts.

Teachability Without Code

Your business has unique logic that generic models can't anticipate. Defog lets you align the model to your specific business rules without writing any code. Provide feedback on query outputs, and Defog learns from your corrections. For teams with specific requirements, full model fine-tuning is also available.

Advanced Analytics for Complex Questions

Beyond simple queries, Defog handles multi-step reasoning. You can ask questions requiring statistical analysis, trend identification, or multi-stage data transformations. The model can work through complex analytical problems, breaking them down into the appropriate SQL operations.

Who Is Using Defog

Defog serves teams across industries who need fast, secure access to their data. Here's how different roles are putting it to work.

Marketing and Sales Teams

Marketing teams often struggle to get quick answers about campaign performance across channels. With Defog, you can ask questions like "Which email subject lines had the highest open rate last month?" or "What's the conversion rate by traffic source for Q3?" without waiting for analyst support.

Sales teams use Defog to prioritize leads by asking "Which leads have engaged with our content more than 3 times in the past month?" or "What's the average deal size by industry?" The result: faster decision-making and improved conversion rates.

Manufacturing Operations

Manufacturing teams deal with massive datasets from production lines, supply chains, and quality control systems. Defog helps identify production bottlenecks by asking "Where are we seeing the highest defect rates in the past week?" or "What's our equipment utilization by shift?"

Supply chain managers use it to predict inventory needs: "Based on current demand trends, what should our reorder points be for top 20 SKUs?" The ability to catch issues before they escalate has transformed quality control workflows.

E-commerce and Retail

Retail businesses face constant pressure on pricing and inventory. With Defog, you can analyze demand patterns alongside competitor pricing and cost data to recommend optimal pricing strategies. Questions like "Which products are at risk of stockout based on current velocity?" help prevent lost sales.

Advertising teams ask "What's our ROAS by campaign for the past 30 days?" to allocate budget more effectively.

Financial Services and Investment

Investment teams need fast analysis of market trends and portfolio performance. Defog enables questions like "What's the risk-adjusted return by sector in our portfolio?" or "Show me the correlation between interest rate changes and our fixed income holdings."

Compliance teams use it to generate audit reports and monitor exposures across portfolios—all through natural language queries.

Healthcare and Life Sciences

Clinical research teams use Defog to analyze patient outcomes data and identify patterns across treatment groups. Commercial teams analyze drug market share trends and physician prescribing patterns.

The ability to quickly query clinical trial data has accelerated research insights, while market analysis capabilities help identify new commercial opportunities.

Technical Architecture and Deployment

Understanding the technical foundation helps you make informed decisions about integration and security.

The SQLCoder Model Family

SQLCoder is Defog's proprietary large language model, specifically fine-tuned for Text-to-SQL tasks. Unlike general-purpose LLMs that have been adapted for database work, SQLCoder was trained from the ground up on SQL generation tasks.

The model family includes three sizes:

• SQLCoder-8b: Lightweight and fast, ideal for real-time queries and smaller datasets
• SQLCoder-14b: The balanced option for most enterprise workloads
• SQLCoder-32b: Maximum capability for complex queries and challenging schema structures

You choose which model to deploy based on your performance requirements and infrastructure capacity.

SQLEval: Ensuring Accuracy

Defog maintains SQLEval, an open-source evaluation framework for Text-to-SQL models. This ensures transparent, reproducible benchmarking so you can verify model performance on your specific use cases. It's also how Defog continuously measures and improves SQLCoder against industry benchmarks.

Deployment Options Explained

Cloud Hosted — The fastest path to value. Deploy via AWS or GCP Marketplace with one-click setup. Everything runs on Defog's infrastructure, with 5-minute setup time. Ideal for teams wanting quick adoption without infrastructure management.

Self-Hosted — Complete control over your data. Deploy Docker containers on your own infrastructure—on-premises, in your private cloud, or within your VPC. This option supports the largest model (32b) and offers unlimited query volumes. Best for organizations with strict data residency requirements or existing Kubernetes infrastructure.

Desktop Application — For individual analysts who want a local experience. Download the app, connect to your databases, and query directly from your workstation.

Slack Integration — Bring data queries directly into your team conversations. @Defog in any channel to ask questions and get results shared with the entire team. This has proven particularly popular for marketing and operations teams who already live in Slack.

Security and Compliance

Defog includes deterministic security filters that prevent the model from generating queries that could expose sensitive data or perform unauthorized operations. Enterprise features include SSO integration (SAML/OIDC), role-based access control, and comprehensive audit logging.

For organizations requiring formal agreements, Defog offers MSA (Master Service Agreement) and DPA (Data Processing Agreement) to meet enterprise procurement requirements.

Performance at Scale

The Enterprise Cloud tier supports up to 200 queries per minute—enough for demanding analytical workloads. Self-hosted deployments have no rate limits, giving you unlimited query capacity.

Pricing Plans

Defog offers three tiers to match different team sizes and requirements.

Open Source is ideal for technical teams wanting to evaluate SQLCoder capabilities or organizations with strong engineering resources who prefer self-managed deployments. You download the model directly from GitHub or Hugging Face and run it on your infrastructure.

Enterprise Cloud suits teams that want rapid deployment without infrastructure management. At $5,000/month for 20,000+ queries, it works out to $0.25 per query—often a fraction of the cost of analyst time for equivalent requests. The white-glove onboarding gets your team productive within days, not weeks.

Enterprise Self-hosted is for organizations requiring maximum control, compliance flexibility, or unlimited query volumes. Annual contract pricing includes full support and custom deployment assistance.

Frequently Asked Questions

---

Ready to put your data to work? Visit defog.ai to get started, explore documentation at docs.defog.ai, or reach the team at founders@defog.ai for personalized guidance on your use case.