About the role
Every day, millions of people skip Google and ask an AI for answers instead. The AI picks a handful of brands to recommend. Most companies have no idea if they made the list. PingAura fixes that — our flagship product is the AI Coworker, an agent that does the optimization work for them.
Building real AI agents is a different discipline from building software. Agents are nondeterministic. They reason in steps. They call tools. They fail in interesting ways. Evals matter more than test coverage. The model layer changes every quarter, and the architecture has to absorb that without breaking.
We are hiring our first dedicated AI Agent Engineer to own this layer. You will design the agent loops, the tool surfaces, the eval bench, and the model-routing logic behind the Coworker.
We are pre-seed and backed by 14 CXO angel investors, Google for Startups, and AWS. The product is live. The customers are paying.
Responsibilities
- Design and iterate on agent behaviors across real-world AEO tasks and long-horizon workflows
- Own the Coworker's reasoning loop, tool calls, and structured outputs across OpenAI, Gemini, and Anthropic
- Build the eval bench. Measure quality, latency, cost, and regression across model versions. Make evals the gate on shipping
- Improve agent performance through prompting, tool-use strategies, context construction, and model-facing experimentation
- Design and build multi-model orchestration — routing, fallbacks, and cost optimization across providers
- Build guardrails: input validation, output filtering, hallucination detection, and rate management
- Instrument agent pipelines with observability — tracing, token tracking, and latency monitoring on Langfuse
- Turn improvements into measurable gains in solve rate, usefulness, and economic value for customers
You may be a good fit if
- You have 2+ years of production experience building with large language models — not demos or hackathon projects
- You have shipped agentic systems, prompt chains, or RAG pipelines that real users depend on
- You can explain in one sentence why an AI agent is not a chatbot
- You have a deep understanding of tokenization, context windows, structured outputs, tool use, and multi-model tradeoffs
- You write clean, testable Python or TypeScript
- You think about cost per call as much as you think about accuracy
- You have dealt with the hard parts: hallucinations, latency spikes, provider outages, and meaningful evals
- You have built an eval bench before, or have felt the pain of not having one
Strong candidates may also have
- Experience with eval frameworks: Langfuse, Braintrust, RAGAS, DeepEval, OpenAI Evals, or comparable
- Experience designing reasoning systems for multi-step, tool-using agents
- Background in search, content, or marketing technology
- Open-source contributions to AI tooling, agent frameworks, or developer infrastructure
What we work with
- Language: TypeScript across the stack; Python for ML and eval tooling
- Web: Next.js 16 (App Router), React 19, Server Actions, Tailwind, Shadcn UI
- Database: PostgreSQL on Supabase with row-level security; pg_cron and pgmq for scheduled and queued work
- Cache and rate limiting: Redis on Memorystore — caching, distributed rate limiting, queue patterns
- AI: OpenAI, Gemini, and Anthropic via a multi-provider routing layer
- Observability: Langfuse for LLM traces, Sentry for errors, plus standard cloud logging and monitoring
- Cloud: GCP for compute, data, and storage; AWS for CDN
- Workflow: Turborepo monorepo, pnpm, Cursor and Claude Code daily
Compensation
- Competitive salary — you're joining at the ground floor
- Early team means you help shape the culture, the stack, and the way we work
Why this team
You own the AI systems, not inherit them. There is no legacy architecture — you define it. The product is live and the customers are paying.