Software Engineer — AI-Native Full Stack

Software Engineer — AI-Native Full Stack Bolo.ai Bay Area (Hybrid) | Salt Lake City Area (Remote) | Full-Time Senior Engineer The Role Has Changed Three person engineering teams are building what used to take thirty. Not by working harder, but by working differently. The engineers shipping at this pace don't write code. They write specs precise enough that agents implement them correctly. They build harnesses. CI gates, structural tests, linting rules, and architectural enforcement that mechanically prevent entire classes of agent mistakes. They design validation systems where agents write the tests and humans verify that features actually work from the user's perspective. The code is a generated artifact. The spec, the harness, and the validation infrastructure are what engineers maintain. This is how we work at Bolo.ai. We're hiring engineers who already work this way, or who have the depth to start. The Company Bolo.ai builds generative AI systems for the energy industry, making daily work faster, safer, and better for heavy industry workers. We have Fortune 500 contracts, production deployments, and growing enterprise demand. We're scaling. Energy adds real constraints. Regulatory compliance, data residency, operational technology integration, deployment across cloud and on-premises infrastructure. These constraints make the architecture harder and the work more interesting. The Work You'll spend your time on four things: Specifications. You write behavioral specs, architectural constraints, and feature requirements that agents implement against. When agent output misses the mark, you tighten the spec. Not by adding more words, but by being more precise about what "correct" means. This requires understanding the system deeply enough to define its behavior at every layer. Harness. You build and maintain the infrastructure that keeps agents producing reliable code. Structural tests that enforce architectural boundaries. Linting rules where every failure message teaches the agent what went wrong. CI gates that reject drift. Structured knowledge bases agents can navigate. The principle: every class of agent mistake gets a mechanical fix so it never recurs. Validation. Agents write the code. Agents write the tests. You verify that features work from the user's perspective, under real deployment conditions, against edge cases that matter in production. You define scenarios and acceptance criteria. You build the end-to-end checks, behavioral verification, and automation that make this trustworthy at scale. When something breaks, your job is diagnosing whether the failure is in the spec, the harness, or the agent's implementation, and fixing the right layer. Architecture and operations. Our systems run across cloud providers and on-premises environments. You design modular abstractions, clean interfaces where deployment targets don't leak into application logic. You own production systems used by energy companies in regulated environments where failures have real consequences. Reliability, observability, and graceful degradation matter here. What Makes Someone Good at This 7+ years of engineering experience, applied at a higher altitude. You need years of building and debugging production systems. Not because you'll write every line, but because you can't design a harness that catches real failures, write a spec that anticipates edge cases, or diagnose a broken feature across the full stack without that foundation. The depth serves the abstraction. Systems thinking over code fluency. How components interact. Where failures cascade. What breaks when requirements change. What to anticipate before it happens. This is what agents are worst at and what matters most. An agent-driven workflow. You already direct AI agents (Claude Code, Codex, Cursor, or similar) to handle implementation while you focus on architecture, specification, and validation. Or you have the engineering judgment to make that transition and the motivation to do it