Senior Applied AI Evaluation Engineer

Posted: 2026-08-03

Parallel Wireless is reimagining mobile networks with innovative, energy-efficient Open RAN solutions. Join us as we lead the future of telecommunications, driving innovation through green and sustainable networks. Learn more about our mission, vision and values.

We are looking for a hands-on Senior Applied AI Evaluation Engineer to improve how our RAN R&D organization analyzes engineering tickets, supports Root-Cause Analysis, recommends ownership, and learns from resolved cases. You will build rigorous evaluation datasets, establish meaningful baselines, compare internal and approved external tools, analyze failure modes, and prototype improvements across retrieval, classification, prompting, agent workflows, and model selection. This role is focused on measurable, evidence-based improvement rather than AI demonstrations. You will assess whether AI-generated conclusions are accurate, grounded in evidence, appropriately calibrated, and useful to engineering teams. As a secondary area of focus, you will analyze engineering workflows at case and team level to identify bottlenecks, handoffs, dependencies, and opportunities for process improvement.

- Define high-value RAN ticket-intelligence use cases, acceptance criteria, evaluation metrics, and quality guardrails.

- Build and maintain representative, versioned evaluation datasets using resolved tickets, Root-Cause Analysis, logs, test evidence, code changes, reviews, reassignment history, and outcomes.

- Establish current-tool and non-AI baselines before evaluating new LLM, RAG, search, or agent-based approaches.

- Evaluate approved internal, commercial, local, and open-source solutions using secure and reproducible data-handling processes.

- Measure retrieval quality, groundedness, diagnosis accuracy, citation support, routing recommendations, calibration, abstention, latency, cost, and human effort.

- Design held-out, time-based, edge, and adversarial test cases while preventing data leakage and future-outcome contamination.

- Analyze failures and turn incorrect conclusions, misrouting, unsupported claims, and missed evidence into prioritized improvements.

- Prototype improvements in search, metadata, context construction, prompting, reranking, classification, agent workflows, and model selection.