AI/ML Engineer Intern - Advanced Track (Summer/Fall 2026)

Venturseed · New York City · 3 weeks ago

Seniority: Intern
Employment: Internship
Category: ML ai

agentic-ai ai anthropic api autogen automation ci-cd cloud crewai dart docker embeddings

### About Us We are building an AI program across our products, internal tools, and customer-facing workflows. The work includes AI agents, model routing, prompt evaluation, embeddings, retrieval, product copilots, website bots, and cost-controlled GPU inference. Our platform is GCP-first, with AI services routed through our own gateway and external GPU providers used where they make sense for cost and flexibility. The goal is not to experiment with AI for its own sake. The goal is to build reliable, useful AI systems that improve product experience, automate internal work, and support future SaaS revenue. We are looking for a technically strong AI intern with hands-on AI project experience who can help prototype, evaluate, and improve real AI features. You will work across prompts, data, APIs, model behavior, and deployment constraints. --- ### Why Join Us - **Real product impact:** Work on AI features intended for real users and internal business workflows. - **Practical AI stack:** Build with gateways, APIs, embeddings, evaluations, GPU providers, and cost controls. - **Senior technical mentorship:** Work alongside engineers and founders building production SaaS products. - **Startup exposure:** See how AI decisions connect to product, infrastructure, growth, and revenue. - **Trial-to-hire path:** Designed as a pipeline for permanent AI, engineering, or product roles in 2026. --- ### What You’ll Do - Prototype AI agents, copilots, chat workflows, and automation tools for our products and internal operations. - Build and test prompt flows for tasks such as summarization, classification, recommendations, support, research, and workflow automation. - Work with embeddings, retrieval, and structured context to improve answer quality and reduce hallucinations. - Evaluate model outputs using test cases, rubrics, regression checks, and human review workflows. - Help compare model providers and GPU-backed inference options based on quality, latency, reliability, and cost. - Support AI gateway workflows, including routing, usage tracking, tenant controls, and budget-aware inference. - Create small tools, scripts, and dashboards that help monitor AI behavior, usage, cost, and quality. - Document prompts, evaluation results, model assumptions, and internal AI playbooks. --- ### What You’ll Bring - Hands-on experience building AI agents, chatbots, RAG systems, copilots, automation workflows, ML prototypes, or AI-powered applications. - Ability to write clear Python, JavaScript/TypeScript, Dart, or similar code for prototypes, integrations, and automation. - Understanding of LLM basics, prompting, embeddings, retrieval, agents, or model evaluation. - Experience working with at least one AI API, open-source model, vector database, model-serving tool, or automation framework. - Comfort testing AI outputs critically rather than assuming a model response is correct. - Ability to break ambiguous AI ideas into concrete experiments and measurable results. - Interest in production constraints such as cost, latency, reliability, privacy, and user experience. - Clear communication and documentation habits. --- ### What You’ll Learn & Build On - How to move from AI prototype to production-ready product workflow. - How AI gateways, provider routing, usage limits, and budget controls work. - How to evaluate LLM outputs with repeatable tests instead of subjective impressions. - How embeddings and retrieval improve AI workflows. - How startups decide which AI features are worth building, shipping, or cutting. - How to balance model quality, infrastructure cost, and user experience. --- ### Nice to Haves - Portfolio projects, GitHub repos, demos, hackathon projects, technical blog posts, or shipped AI features. - Familiarity with OpenAI, Anthropic, Gemini, open-source LLMs, vLLM, Ollama, LangChain, LlamaIndex, CrewAI, AutoGen, or similar tools. - Experience with vector databases, embeddings, semantic search, document processing, or retrieval pipelines. - Exposure to cloud platforms, APIs, Git, Docker, Firebase, GCP, Kubernetes, or CI/CD. - Interest in GPU infrastructure, model serving, inference optimization, or AI cost controls. - Startup, internship, freelance, research lab, or product-building experience. --- ### Compensation This is a paid internship program with a flat completion stipend. - **Program compensation:** Flat stipend paid at the end of the program. - **Duration:** Fall 2026 internship program, with the option to extend based on performance and business needs. - **Future role potential:** Strong participants may be considered for future full-time, part-time, or contract roles in 2026. - **Equity:** Not offered for the internship program, but may be considered for future permanent roles depending on role scope and company stage. Compensation is structured for an early-stage AI program internship. The role is best suited for someone who already has hands-on AI project experience and wants to apply it to practical AI systems with real product and business use cases. --- # Candidate Screening Questions **1. AI Project Experience** Provide a live working link of a SAAS product, game, or other app that is currently using your AI or RAG model you built. Please be prepared to showcase this on the interview. What problem did it solve, what tools did you use, and what did you personally implement? **2. AI Workflow Design** Pick a business workflow that could benefit from an AI assistant. How would you design the first prototype, and how would you know if it is useful? **3. Prompting and Evaluation** How would you test whether a prompt is producing reliable answers over time? What kinds of examples or edge cases would you include? **4. Retrieval / Context** When would you use retrieval or embeddings instead of putting everything directly into a prompt? What risks would you watch for? **5. Model Comparison** If two models give different answers for the same task, how would you decide which one is better? **6. Cost Control** How would you prevent runaway AI or GPU usage while still allowing useful experimentation and testing? **7. Production Readiness** What would need to be true before you would let an AI feature be used by real customers? **8. Debugging AI Behavior** If an AI assistant gives a confident but wrong answer, how would you investigate the cause and reduce the chance of it happening again?

Explore more jobs