The posting
If you've ever dreamt of being there when it all started, this one's for you.
At Zeroport, we're rethinking how humans and AI agents securely access and interact with critical systems.
As AI becomes more autonomous, controlling what it can access is no longer enough. We're pioneering Least Agency: limiting what AI agents (and humans) can actually do, and enforcing those boundaries independently through physical containment.
Our approach combines purpose-built hardware, robust software, and non-IP connectivity to enable secure, real-time work for both AI agents and humans, without exposing the protected network.
It sounds bold, we know. That's exactly the point.
Who We Are
Zeroport is pioneering Least Agency for AI and human network access.
Our hardware-enforced architecture creates physical boundaries between AI agents, human users, and critical systems, limiting what they can reach and do by design.
We're a focused, technical team with a strong builder mentality, moving fast to redefine how humans and AI securely interact with the world's most critical systems.
About the Role
We're hiring a senior AI Engineer to join our AI team building our autonomous AI agent and helping extend AI capabilities across the company.
The agent runs on NVIDIA hardware integrated inside our purpose-built platform, watches up to 200 concurrent video streams from our hardware-isolated remote access solution, and is the only path to compliance and threat detection in an environment that intentionally blocks every traditional monitoring channel. Larger deployments span multiple edge units coordinating across sites, and the whole system ships into fully air-gapped customer networks — updating, learning, and evolving without ever touching the cloud.
This is not a wrap-a-model-in-an-API role. You'll work on real production-grade agent systems, deep multimodal inference at the edge, and the infrastructure that makes both ship and scale.
What You'll Do
- Design and ship our agent systems end-to-end — perception, reasoning, memory, retrieval, and the loops that connect them. Production agents, not prototypes.
- Optimize multimodal inference for real-time operation at the edge — model choice, quantization, and batching to hit our latency and concurrency budgets.
- Design embedding pipelines, vector storage, and hybrid retrieval that power the agent's search, behavioral analytics, and rule generation.
- Architect how multiple edge units coordinate at scale — sharing context, correlating activity, and behaving as one coherent system for our largest deployments.
- Build the agent's air-gapped lifecycle — updating, learning, and evolving entirely inside customer private networks with no cloud connectivity.
- Run focused research on new open-source models and inference frameworks, and bring back insights and prototypes that inform our roadmap.
Plus occasional cross-company AI projects across the rest of the company.
Requirements
Who You Are
We care about a particular mindset more than any specific item on a checklist. The person who'll thrive here is someone already living inside the modern AI stack - not planning to start. You read model release notes the way other people read the news. You've built real things with new tools the week they came out. Agentic dev workflows aren't something you've heard about; they're how you already work. When a new model drops, your first instinct is to put it on the bench. And underneath all of it, you're a builder - you ship, you write clean software, and you think about latency, cost, and what the user actually needs.
What We're Looking For
- 6+ years of software engineering experience, with the last few focused on AI / ML systems
- Deep, hands-on experience designing and shipping agent systems in production - not demos. You understand agent architectures, memory, tool use, evaluation, and the failure modes that matter at scale.
- Strong fundamentals in embeddings, RAG, and hybrid retrieval, with real experience designing vector storage and retrieval pipelines for production use
- Deep production backend foundations - strong Postgres knowledge (functions, triggers, indexes, extensions, atomic operations, and similar depth), real-time client channels (WebSockets, SSE), async and event-driven backbones (pub/sub, queues, background tasks, webhooks), the ability to design systems and features around multithreading and multiprocessing, and dynamic resource management for high-throughput, latency-sensitive workloads.
- Excellent architecture instincts and strong Python - you write code that scales and that other engineers can build on
- Product sensibility - you can reason about trade-offs and what's worth building, not just what's technically possible
Nice to Have
- Experience with vision-language models and multimodal systems
- Experience running AI workloads on edge hardware (Jetson, NVIDIA accelerators, custom inference rigs)
- Familiarity with vLLM, TensorRT, or similar inference stacks; quantization and model optimization
- Experience with video streaming technologies (WebRTC, GStreamer, STUN/TURN)
- Background in networking, cybersecurity, or systems-level Linux
- Experience with model fine-tuning and dataset construction
Why Join Us
- Be part of designing innovative solutions for some of the toughest challenges in network security
- Flexible work setup, minimal bureaucracy
- Real ownership and impact from day one
- A collaborative and innovative work culture
- Competitive salary and benefits
Let's build it together!
Even if you don’t meet every requirement – we’d still love to hear from curious, motivated people.
