The Hog

Agents needed the web, and the web came in twelve formats.

Search, scraping, and enrichment each lived with different providers. Each had its own contract, its own failure modes, and its own idea of a result. An agent that wanted context had to stitch all of it together before doing any real work.

One contract over a waterfall of providers.

I built and launched The Hog's API platform for deep web research and enrichment. Providers sit behind a waterfall: try the best source first, fall through on failure, and return one normalized, agent-ready result.

The endpoints reached 500+ consumers in the first week.

Deep research runs long, so the runtime has to be durable.

A research job can take minutes and touch dozens of sources. I built a durable agent runtime that persists state in Redis and isolates execution in AWS Fargate sandboxes. A job survives restarts, and one tenant's execution never shares a process with another's.

Every execution attributed to a user and an organization.

A production API has to know who ran what. I built a multi-tenant metering and billing system that attributes each API execution by user and organization, so usage, spend, and limits are correct per account.

• The Hog's API platform for deep web research and enrichment, consolidating a waterfall of providers behind production endpoints with agent-ready results.
• A durable agent runtime for deep research with state persisted in Redis and execution isolated in AWS Fargate sandboxes.
• A multi-tenant metering and billing system attributing API executions by user and organization.

First week

500+ consumersOn the launched API platform.

State

RedisDurable across runtime restarts.

Isolation

AWS FargateOne sandbox per execution.

Attribution

User + orgMetered and billed per execution.