Software Engineer (SDE 3) - Agentic Workflows

About Bytebeam
Bytebeam builds connected-vehicle infrastructure and is developing a fleet-management application that brings live telemetry, diagnostics, alerts, actions, and operational data into one working system.

Role Overview

We are looking for a senior engineer to build reliable agentic workflows for fleet operations and internal engineering processes. This is not a prompt-only role. You will design software where models use well-defined tools, retrieve the right operational context, ask for approval when needed, and produce actions that can be evaluated, audited, and trusted.

What You'll Do

• Build agent-assisted workflows that help users investigate vehicle issues, understand fleet state, prepare reports, and complete repetitive operational tasks.
• Design typed tools and service interfaces over telemetry, alerts, actions, documentation, and fleet-domain data.
• Implement orchestration, state management, retrieval, guardrails, retries, and human-in-the-loop approval.
• Create offline and online evaluation suites covering task success, factuality, safety, latency, and cost.
• Instrument traces and feedback loops so failures can be reproduced and improved systematically.
• Defend workflows against prompt injection, over-broad permissions, unsafe tool use, and ambiguous user intent.
• Partner with backend, frontend, QA, product, and domain experts to move prototypes into dependable product experiences.
• Make practical decisions about when deterministic software is better than a model.

Must-Have Skills

• 5+ years of software engineering experience building production backend or full-stack systems.
• Strong proficiency in Rust or Go.
• Experience integrating LLM APIs or building tool-using, retrieval-augmented, or workflow-oriented AI systems.
• Solid understanding of API design, databases, asynchronous work, testing, observability, and production reliability.
• Ability to design evaluation datasets and diagnose failures beyond subjective demos.
• Clear written communication and strong product judgment.

Good-to-Have

• Experience with fleet operations, IoT, telematics, customer support, or incident-investigation workflows.
• Familiarity with vector/search systems, structured generation, workflow engines, or model routing.
• Experience with permission models, audit logs, PII controls, red teaming, or AI security.
• Contributions to developer tools, AI infrastructure, or evaluation frameworks.

What Success Looks Like

You ship a small set of high-value workflows that users repeatedly trust, with measurable task success, controlled permissions, predictable cost and latency, and a clear path from every production failure to an improved test.