SUREVoter. Let’s Fix the Code of Democracy.
THE PROBLEM The system is broken. Princeton study proved that the average American citizen has a near-zero impact on public policy.
THE MISSION We are building GovTech 2.0. We are using AI to make the government efficient, transparent, and responsive. We are building the tools to parse the complexity of legislation, expose the incentives of politicians, and give the American voter a voice that can finally compete with big money.
Why This Role Exists
SureVoter is building an AI-native civic technology platform. The AI layer has to be trustworthy enough that voters rely on it, and auditable enough that regulators can get the data they need. This role owns the reliability, safety, and controllability of that system from day one.
How We Work
Every person at SureVoter uses AI daily as a core part of how they think, build, and ship. We use AI to write code, research problems, synthesize information, draft documents, evaluate outputs, and accelerate decisions. If you're someone who has been actively adopting AI tools into your workflow and pushing the boundaries of what's possible with them, you'll fit right in. If you're waiting for AI to prove itself before you engage, this isn't the right environment. We're building an AI-native product and we operate as an AI-native team.
What You'll Do
Core AI Systems (70%)
- Design, build, and maintain production AI workflows including retrieval-augmented generation (RAG) pipelines, agentic task systems, and structured reasoning chains
- Build and operate retrieval pipelines that pull from structured and unstructured data sources, normalizing messy real-world inputs into queryable context
- Implement evaluation harnesses that measure hallucination rate, bias drift, factual grounding, and neutrality, and make results auditable
- Build deterministic logging and replayable model output infrastructure so any AI-generated response can be traced back to its source data, prompt, and model version
- Develop and maintain prompt engineering systems, agent workflows, and prompt chains for complex user interactions
- Select and integrate appropriate embedding models and vector databases (Pinecone, Weaviate, pgvector, Qdrant)
- Optimize for latency, cost, and quality across AI API calls
Safety & Trust (20%)
- Implement guardrails that enforce the product's neutrality constraints at the systems level
- Build bias detection and neutrality metrics that run continuously in production
- Design content filtering and moderation systems for AI-generated outputs
- Build model-agnostic architecture so the system can swap underlying models (OpenAI, Anthropic, open-source) without rearchitecting
Infrastructure & Scale (10%)
- Own deployment, monitoring, and incident response for AI services
- Build modular architecture that supports multiple deployment targets
- Monitor AI system performance and iterate based on real-world usage and feedback
What We're Looking For
Required
- 5+ years shipping production backend or ML systems
- Hands-on experience with LLM integration: RAG, prompt engineering, structured output, tool use, and evaluation
- Deep experience with LLM APIs (OpenAI, Anthropic, AWS Bedrock) and orchestration patterns
- Proficiency with vector databases (Pinecone, Weaviate, pgvector, Qdrant, FAISS)
- Strong software engineering fundamentals in Python
- Experience building data pipelines that ingest, normalize, and structure messy real-world data
- Comfortable with cloud infrastructure (AWS preferred)
- Engineering mindset oriented toward auditability, reproducibility, and determinism
- Active user of AI coding tools and always looking for new ways to accelerate
Preferred
- Experience with civic data, government data systems, or public records
- Familiarity with bias evaluation frameworks for language models
- Prior work in high-trust or regulated domains (healthcare, fintech, legal)
- Experience at an early-stage startup
- Exposure to agentic AI patterns (multi-step task decomposition, tool use, autonomous monitoring)
- Familiarity with LangChain, LlamaIndex, or similar orchestration frameworks
- Experience with streaming responses and real-time AI features
What Success Looks Like (First 12 Months)
- Production AI systems are live, answering real user questions with traceable, replayable responses grounded in structured data
- Evaluation infrastructure is running: hallucination rate, bias drift, and neutrality are measured continuously
- The system architecture cleanly enforces data access boundaries across organizational entities
- Users trust the system to help them understand, not to push them in a direction
Why SureVoter
This is an agentic startup building something that actually matters. There's no large team, no layers of management, no legacy system. You'll build the AI foundation of a civic technology platform from scratch, designed to make civic participation more accessible.
We're looking for senior people who have done this before and want to do it again for something bigger. This is a small team moving fast on a hard problem. The pace is startup pace. The commitment is real. If you get energy from building at full speed alongside people who are all-in, this is going to be a lot of fun.
If that sounds like your kind of thing, we should talk.
THE OPPORTUNITY Some startups disrupt taxi cabs and marketing widgets. Others dare to do something more extraordinary. We will disrupt the dysfunction of the Governments to help them run more efficiently - starting with the United States.
Join us. Let’s bring power back to the people.
