Applied AI Engineer (Agentic Workflows & RAG) - Master-Level Internship

Job description

About Us: At Vosyn, we embrace the exciting, game-changing world of Artificial Intelligence, driving innovation and pioneering impactful projects across various industries. We are a trailblazing Language Synthesis AI firm reshaping global communication by dissolving language barriers and empowering users. We believe in fostering a culture of flexibility, continuous improvement, and solution-focused strategies. Here, every idea is welcomed, nurtured, and has the potential to scale to new heights. Currently, we're at the forefront of a significant IPO endeavor, truly a unicorn in the making. We invite you to be part of our journey and leave your imprint on the future of AI. About the Role: We are seeking a sharp, technically deep Applied AI Engineer Intern to own the intelligence layer of what we build. This role is ideal for a Master’s level student who does more than use AI coding tools — you understand how modern language and reasoning models actually behave, and you build with them as components. You will design and ship agentic workflows (systems where a model plans, acts, and re-plans in a loop), build retrieval-augmented generation (RAG) and search integrations that ground AI in real, current data, choose the right model for each job, and write the evaluations that prove the system works rather than just appears to. This is the role that delivers the “AI” in AI consulting: when a client has a scoped roadmap, you are the person who stands the tool up. Tools & Tech Stack: • Agentic coding: Claude Code (primary), plus Cursor or Windsurf • AI APIs & SDKs: Anthropic Claude API and comparable model APIs; reasoning and instruction-tuned models • Retrieval & RAG: vector databases (e.g., pgvector, Pinecone, Weaviate), embeddings, semantic and hybrid search • Orchestration & integration: MCP (Model Context Protocol), function/tool calling, agent frameworks • App & data layers: React.js / Next.js, Node.js / Python, Supabase or Firebase,PostgreSQL • Evaluation: prompt and output evaluation harnesses, test sets, regression checks for non-deterministic systems • Version control & documentation: Git / GitHub, Notion Key Responsibilities: • Design, build, and harden agentic workflows that plan and take actions reliably — and understand why agents fail (context loss, compounding errors, no feedback signal) and how to structure tasks so they succeed. • Build retrieval (RAG/search) pipelines that fetch the right client data and ground model outputs in it, integrated into core applications rather than demos. • Select the right model for each task — reasoning model vs. fast instruction model — and be able to justify the trade-off in latency, cost, and quality. • Engineer prompts and context structures appropriate to the model class, including knowing when reasoning models need framing rather than step-by-step hand-holding. • Write evaluations for AI features, because with non-deterministic models “it worked once” is not evidence that it works. • Connect AI tools to internal systems and data sources via APIs or MCP to power real client use cases. • Review and validate AI-generated code and automated workflows critically for correctness, security, and safety. • Collaborate with the Builder and the Integration & Data Engineer to deliver complete, working solutions, and document workflows, prompts, and integrations in Notion. About You: • Currently enrolled or recently graduated from a Master’s program in Computer Science, Software Engineering, AI/ML, Information Systems, or a related field. Master’s program enrollment or completion is mandatory. • Strong, demonstrable hands-on experience with AI coding assistants and the Claude API or comparable model APIs — portfolio, GitHub, or live examples strongly preferred. • A working understanding of how modern LLMs and reasoning models behave: context windows, the difference between reasoning and instruction models, and when to reach for each. • Practical experience with at least one of: building an agentic workflow, building a RAG/retrieval pipeline, or integrating models via tool calling or MCP. • Excellent prompt-engineering and context-management skills. • Coding fluency in JavaScript/React and/or Python sufficient to build, evaluate, and fix AI-generated output. • An instinct for evaluation: you want to measure whether the AI is actually correct, not just plausible. • Excellent verbal and written communication skills within a cross-functional team environment. New graduates are encouraged to apply. We believe exceptional talent often emerges from diverse paths. If you possess a profound curiosity, a genuine passion for continuous personal and professional growth, and a strong desire to apply your unique abilities to create significant impact within our team, we strongly encourage you to apply even if your background doesn’t align perfectly with every single qualification.