Founding AI Platform Engineering Lead

Eros is building sovereign cultural AI: systems designed to make advanced intelligence culturally relevant, rights-aware, governable, and commercially useful. We are bringing together a major film and media catalog, a new AI platform, and a founding team, with every data and model decision governed by rights, provenance, evidence, and policy controls.

A first gateway, router, and foundation-model release are already in flight. We need a founding engineering lead who can own the architecture, ship production code every week, and raise the level of a small cross-time-zone team.

This is a full-time founding role for someone who wants long-term ownership and remains deeply hands-on.

What you will own

Define and evolve the common platform architecture across model access, inference, model lifecycle, evaluation, and agent-runtime interfaces.

Turn ambiguous product and technical goals into executable boundaries, contracts, milestones, and acceptance tests.

Write and review production code, not only diagrams or strategy documents.

Establish secure multi-tenancy, identity propagation, observability, cost controls, release gates, and failure handling.

Make practical build-versus-buy and open-weight-versus-hosted decisions using quality, latency, cost, security, and portability evidence.

Lead design reviews, incident reviews, and technical hiring panels.

Create a clean handoff path across internal engineers, contractors, and product owners.

You are likely a fit if you have

Eight or more years in backend, platform, distributed-systems, or ML infrastructure engineering.

Staff, principal, founding, or early-engineer ownership of a production distributed system.

Personally built or operated LLM-era infrastructure under real traffic, such as an AI gateway, inference service, evaluation platform, model control plane, or agent platform.

Owned incidents, rollbacks, capacity decisions, observability, and measurable quality, latency, and cost tradeoffs.

Led a small team while remaining deeply hands-on in code and design.

Strong written communication and several hours of reliable overlap with US Eastern time.

Experience with vLLM, TGI, TensorRT-LLM, KServe, Ray Serve, Kubernetes, OpenTelemetry, policy engines, or multi-provider model orchestration is useful, but named production ownership matters more than keyword coverage.

To apply, please address these questions

Name the most relevant production platform you personally owned. What traffic or scale did it serve, what code or architecture did you own, and what measurable outcome changed because of your work?

Describe one serious production incident or failed rollout. What did you diagnose, what did you change, and what permanent control did you add?

How would you separate policy eligibility from model or route selection in a multi-tenant AI platform?

Share one sanitized artifact you can walk through live, such as a repository, code sample, design document, incident review, or system diagram.

How many hours of reliable overlap can you provide with US Eastern time?