Conner Aldrich, Co-Founder & CTO at GovSignals, shares how they run large-scale AI ingestion, proposal generation, and a dual cloud/self-hosted FedRAMP High deployment using Trigger.dev. GovSignals has raised $5.5M and has achieved rapid self-sustained growth, serving contractors from small teams to Fortune 500s with billions in annual awards.
Building secure, high-performance AI for government contracting
Our mission at GovSignals is to help companies find and win government contracts with speed and precision. We help teams discover new funding streams and generate fully compliant, winning proposals in hours instead of months.
We’re also the only AI proposal-generation platform to achieve FedRAMP High. This is an extremely difficult authorization to earn, but it allows our Department of Defense (DoD) customers to process controlled information with us securely, a requirement for operating in the federal space at the highest levels.
Our customers span the full spectrum of government contracting, from emerging small businesses to Fortune 500 enterprises working across the DoD, HHS, state agencies, and local governments.
Teams routinely 3× their GovCon revenue using GovSignals. We built the platform to make that level of scale practical.
The challenges we faced before Trigger.dev
Our ingestion pipelines initially ran on Cloudflare Workers Queues, but as they grew more complex we hit hard limits: weak observability, painful manual retries, no batching or scheduling, and workloads that no longer fit the queue model.
Our SvelteKit frontend on Cloudflare Workers had similar issues. Long-running AI tasks like proposal strategy or multi-step generation would routinely time-out, exceed compute quotas, or hit subrequest caps.
What we really needed was a system designed for long-running AI workflows with reliable concurrency control, strong observability, and a clean TypeScript developer experience.
Why we chose Trigger.dev
I discovered Trigger.dev in early 2024 when v3 was announced, and it was immediately clear the team was building what we needed. Trigger.dev gave us:
- Open source and self-hostability (required for FedRAMP High)
- Durable functions with retries and scheduling
- First-class observability across all of our environments
- A lightweight, TypeScript-native API
- A local dev environment that mirrors previews, staging, and production
Open source was essential: we were already planning ahead for a FedRAMP High self-hosted deployment, so everything we adopted needed to be auditable and deployable in extremely controlled environments.
How Trigger.dev fits into our architecture
Trigger.dev powers two core categories of workloads for us:
1. Large background AI processing queues
We process over 10,000 government data sources, from small cities in Ohio to US embassies, statewide agencies, and federal systems like Congress.gov.
A typical pipeline:
- Receive and normalize a government opportunity
- Run AI reasoning models to classify or extract details
- Generate recommendations for customers
- Push structured results back into internal systems
2. Real-time high-compute AI proposal generation
We run parallel AI tasks for:
- Proposal strategy
- Compliance mapping
- Drafting structured proposal content
Strict concurrency limits are essential because the sources we crawl and the APIs we interact with often have unusual or sensitive rate patterns.
Other key features we rely on:
- Queues and concurrency controls for complex ingestion
- Manual triggering + Slack alerts to retry edge cases quickly
- Full observability for payloads, traces, timing, retries, and concurrency across every environment
Today our favorite feature has to be previews, it enables a really seamless experience for our team with a custom preview branch for every github branch tied to every linear ticket. eng-100-we-love-trigger
Operating in both cloud and FedRAMP High
We run a dual-deployment model:
- Commercial SaaS for general customers
- Self-hosted Trigger.dev inside AWS GovCloud for FedRAMP High workloads
FedRAMP High is the highest level of US federal cloud authorization for unclassified systems. It’s required for handling controlled defense information and comes with very strict security and infrastructure requirements.
To self-host Trigger.dev in that environment, we:
- Forked the Trigger CLI to run within restricted networks
- Split task building and registration into independent steps
- Created a custom
trigger-register-tasksimage - Worked closely with the Trigger team during setup
Despite the compliance complexity, deploying Trigger.dev into GovCloud Kubernetes was much easier than expected.
Self-hosting Trigger.dev lets our product team ship the same workflows to both environments without thinking about infra differences. That’s huge for us.
Impact
Trigger.dev is now a foundational part of our platform:
- Ingestion scaled to over 10,000 government sources
- High-compute AI proposal generation runs reliably in parallel
- Observability gives us full visibility across dev → preview → staging → production
- Self-hosting enables us to meet FedRAMP High requirements while keeping a unified developer workflow
Trigger.dev handles our orchestration so we can stay focused on building world-class AI systems for the government contracting ecosystem.
