AI Engineer

We are hiring an AI Engineer to help build our AI platform and the agents that run on it. You’ve spent 2-5 years as a software engineer building real systems, and in the last year or so you’ve gone hands-on with AI: agents, RAG, LLM-powered features. Not just reading about them.

We’re not looking for years of GenAI experience. Nobody has that. We want solid engineering fundamentals with recent, real AI work on top, because the hard problems in AI right now are engineering problems: reliability, cost, latency, security, and testing systems that don’t give the same answer twice.

Most of your work is on the platform: the LLM gateway, the agent runtime, MCP servers, the registry, evals, and observability. You’ll own features inside these systems end to end, with a tech lead and an architect making the deepest design calls and reviewing yours. You’ll also work with our enterprise customer, usually alongside a senior engineer, building inside their environment and their security constraints.

Your code will run inside a bank’s own cloud account. It has to work with no internet egress, with an audit trail on every model call, and with a security team signing off before anything goes live. If that sounds interesting rather than annoying, you’ll do well here.

This is an IC role. You lead by building.

What You’ll Do

• Build features in the LLM gateway: routing, caching, rate limits, per-customer model rules, and cost tracking.
• Build parts of the agent runtime: how agents start, run, hold state, and get versioned and rolled back.
• Build MCP servers for tools and data, hosted inside the customer’s boundary, with scopes and security handled properly.
• Build and extend the AI registry: models, agents, prompts, tools, and datasets, with owners, versions, and approval state.
• Add governance pieces: policy checks, guardrails, approval steps, and audit logs.
• Write evals and regression tests so AI changes ship with evidence.
• Add tracing, cost and latency dashboards, and quality monitoring to what you build.
• Build retrieval pipelines (chunking, embeddings, hybrid search) that respect the caller’s permissions, and measure whether they work.
• Deploy into the customer’s AWS environment: private VPCs, restricted networks, private endpoints. Reproducible, no manual steps.
• Support demos, working sessions, and security reviews with the customer’s engineers.
• Own your features from design through deployment, and operate what you build.

• 2-5 years of software engineering experience shipping production systems: backend, platform, or full stack.
• Strong Python and the habits of a production engineer: testing, code review, observability.
• Good grounding in APIs, data, and distributed systems.
• Working knowledge of AWS (ECS or EKS, Lambda, IAM, networking) and some exposure to infrastructure-as-code (Terraform or CDK).
• Real AI work from roughly the last year: you’ve built agents, RAG pipelines, or LLM applications hands-on and can walk us through what you built and what broke. Production work, serious side projects, and open source all count.
• Experience with the major model APIs (Anthropic, OpenAI, Bedrock) and at least one agent framework (LangGraph, CrewAI, or comparable).
• Practical understanding of LLM behavior: context management, token costs, latency trade-offs, and how to test something that isn’t deterministic.
• Familiarity with RAG and with MCP.
• You want to work directly with enterprise customers and can explain a trade-off in plain language.
• You’re curious about how large organizations work, including their constraints and review processes. Not put off by them., * Work authorization: You must be authorized to work in the United States. No immediate H-1B transfer. If you are on OPT, we can apply for H-1B next year. OPT candidates should have prior experience (in India or the US), for example OPT plus 2-3 years of prior work.

Moring.ai builds enterprise AI that runs in production for regulated industries: financial services and other compliance-heavy organizations. We ship AI applications and agents inside our customers’ own cloud environments, wired into the systems they already run. Not slideware, not demos.