Term: Summer 2026 (10–12 weeks, full-time)

This Position Reports To: Chief Information Officer (CIO)

Key Partners: CIO; Engineering, Data, Front Office Operations, and Customer Success teams working on CORE and Roam

Our Vision, Purpose, and Values

Our vision is to deliver a seamless, personalized travel experience by blending smart technology with the power of human connection. Our purpose is to Stand-Up, Stand-Out — to do something every day that makes the next day better for all. We value creativity, respectful collaboration, the drive to get up and do it, openness to growth, doing what we say, loving what we do, and having fun along the way.

Role Overview

The Engineering Intern, Agentic Systems will join the team building Roam AI, the agentic layer that sits across Atlas’s booking, servicing, and account management workflows. The intern will ship real code into systems that travel agents, account managers, and customers interact with every day — work that directly shapes how Atlas blends human expertise with AI.

The internship is project-based with a defined deliverable, but the intern will operate as a full member of the AI Engineering team: writing production-quality code, participating in design reviews, presenting work to the CIO and Corporate Executive Committee, and collaborating with operators across the business to ground the work in real customer outcomes.

What Success Looks Like in This Role

By the end of the summer, the intern will have:

• Shipped several production-quality features into Roam AI, and CORE with documented impact on a target metric (response quality, deflection rate, time-to-answer, or adoption).
• Built and demonstrated an internal tool or agent that meaningfully improves a defined Atlas business process, with adoption from at least one operating team.
• Contributed to the data and retrieval layer that lets CORE reason over Atlas’s operational data — including evaluation harnesses that measure factuality and grounding.
• Presented findings, trade-offs, and recommendations to the Engineering team and at least one cross-functional partner group (Operations or Program Management).
• Left behind clean, documented, reviewed code; written runbooks; and an end-of-internship write-up that the team can build on after the intern departs.

Responsibilities:

Humanizing Agent Responses

• Improve the tone, structure, and conversational behavior of Roam’s responses so they feel natural, helpful, and unmistakably Atlas — not robotic, not generic.
• Build prompt, system message, and post-processing patterns that adapt to the channel (chat, email, voice transcript) and to the customer relationship (frontline service vs. Tier 4 Travel Executive context).
• Stand up evaluation pipelines — automated and human-in-the-loop — that score responses on clarity, warmth, accuracy, and brand alignment, and use the results to drive measurable iteration.
• Partner with Front Office Operations and Customer Success to gather real conversation samples, identify failure modes, and close the loop on quality.

Agentic Systems for Business Processes

• Identify high-friction internal workflows — implementation onboarding, QBR prep, exception handling, supplier reconciliation, ticket triage — and prototype agentic solutions that remove rote work.
• Build, integrate, and test tools that agents can call: CRM lookups, booking system queries, policy checks, ticket creation, and document generation.
• Design guardrails, approval steps, and audit trails so business-process agents are safe to deploy in regulated and customer-facing contexts.
• Instrument every workflow so adoption, accuracy, and time-saved are measurable from day one.

Data Understanding & Proactive Insight

• Help Roam reason over Atlas’s operational data — bookings, savings and policy compliance, customer health, supplier performance — so people can ask meaningful questions in natural language and trust the answers.
• Contribute to the semantic layer, retrieval, and grounding strategy: schema understanding, metric definitions, joins, and evaluation against known-good answers.
• Prototype proactive surfaces — agents that notice a renewal risk, a budget overage, or an unusual booking pattern and surface it before a human has to ask.
• Work alongside the data team to ensure outputs are accurate, well-cited, and resilient to schema changes.

Engineering Craft

• Write clean, well-tested code in the team’s primary stack and follow code review, branching, and deployment standards.
• Participate in sprint planning, standups, design reviews, and retros as a full team member.
• Document what you build — for the next intern, for the team, and for the operators who will run it.

Skills & Qualifications

• Currently pursuing a Bachelor’s or Master’s degree in Computer Science, Software Engineering, Data Science, or a related field, with an expected graduation between December 2026 and June 2028.
• Strong programming fundamentals in Java and/or TypeScript; comfortable working in a modern codebase with git, code review, and CI.
• Hands-on experience with LLM APIs (Open AI, Anthropic, or equivalent) through coursework, side projects, prior internships, or open-source contributions — including prompt design, tool/function calling, and basic evaluation.
• Working understanding of how retrieval-augmented generation, embeddings, and vector search are used to ground LLM responses in proprietary data.
• Comfort reading and writing SQL well enough to explore an unfamiliar database and validate that an agent’s answer is actually correct.
• Clear written and verbal communication; able to explain technical trade-offs to a non-technical operator and ask for the context they need to do the work well.
• Curiosity about how a real business actually runs — travel, services, operations — and willingness to spend time with the people doing the work.
• Bias to ship: comfortable scoping a problem, making a call, getting something working, and iterating from there.

Nice to Have

• Prior internship or research experience building agentic systems, chat assistants, copilots, or AI-powered internal tools.
• Exposure to agent frameworks (LangGraph, LlamaIndex, CrewAI, Anthropic Agent SDK, or similar) and to evaluation tooling (LangSmith, Braintrust, custom eval harnesses).
• Experience with full-stack web work (React/Next.js, FastAPI/Node) sufficient to ship a usable internal tool end-to-end.
• Familiarity with travel, expense, or B2B SaaS data models — or genuine interest in learning a domain quickly.

Core Competencies

• Engineering craft and code quality
• Practical AI/ML judgment and intellectual honesty about model limitations
• Customer and operator empathy
• Clear written communication and crisp problem framing
• Curiosity, ownership, and bias to action
• Collaboration across cross functional teams

Education & Experience

• Active enrollment in an undergraduate or graduate program in Computer Science or a closely related field.
• At least one prior project or internship that demonstrates the ability to take a software idea from problem statement to working, deployed code.
• Demonstrable interest in applied AI — through projects, coursework, publications, or open-source contributions.

Tools & Technical Skills

• Languages — Strong working proficiency in Java and/or TypeScript; familiarity with one other language is a plus.
• LLM Tooling — Direct experience calling LLM APIs, designing prompts and tool schemas, and reasoning about cost, latency, and safety trade-offs.
• Retrieval & Data — Comfort with embeddings, vector stores (pgvector, Pinecone, or similar), and basic RAG patterns; SQL fluency sufficient to explore Mysql, BigQuery, or Postgres.
• Cloud & Deployment — Working knowledge of at least one major cloud (GCP, AWS, or Azure); familiarity with containers and CI/CD pipelines.
• Collaboration — Git/GitHub, Jira or Linear, Slack, and Google Workspace as the day-to-day operating environment.
• Evaluation — Exposure to eval frameworks, A/B testing patterns, or any disciplined approach to measuring whether an AI feature actually got better.

Internship Logistics

• Term: 10–12 weeks, full-time, summer 2026.
• Location: Remote
• Compensation: Paid internship; rate set by program band and degree level.
• Mentorship: Paired with a senior engineers on the team and a non-engineering partner from the operating business.
• Outcome: Strong-performing interns will be considered for return offers, including post-graduation full-time roles on the Engineering team.