Intern (AI Agents)

We are a machine learning tools development team at Sberbank's Center for Practical Artificial Intelligence. Our goal is to create the best solutions for automating work with structured data: from open-source libraries to AI agents using SotA practices and tools.

We are looking for an intern engineer who can delve deeply into the subject area and take an active part in developing agent systems for working with banking data — from idea to implementation in the business processes of the largest fintech company.

The first stage of selection for this vacancy is communication with an AI recruiter. After applying, expect a message from them on Telegram; the conversation will take approximately 10 minutes. The AI recruiter's task is to clarify missing details and speed up the review of your application. The AI recruiter is just starting their journey, so we ask for your understanding. Your experience and participation will help make it convenient and useful for everyone!

Tasks:

• Develop PoC and MVP AI agents based on LLM using frameworks (LangChain, GigaChain)
• Analyze and adapt existing SOTA practices
• Participate in the full development and integration cycle: from creating prototypes to production deployment and monitoring solutions
• Interact with applied and business teams within the Sber ecosystem.

Requirements:

• Currently studying in the final years of a Bachelor's, Master's, or PhD program at a technical university with a strong background in mathematics and algorithms
• Proficient in Python (NumPy, Pandas, PyTorch, OOP basics, etc.)
• Deep understanding of DL and classical ML
• Practical experience with LLM, RAG, and relevant frameworks and databases (LangChain, FAISS, Qdrant) and a desire to develop in this area
• Experience with development tools (bash, git), including team development experience or a desire to quickly acquire these skills
• Ability to understand scientific publications and modern SOTA solutions, formulate and promote your own research hypotheses.

Will be a plus:

• Publications in A*/Q1 conferences
• Participation in competitions (Kaggle).

Conditions

• Paid internship
• Internship duration: 3 months, 40 hours per week or 6 months, 20 hours per week, hours - from 9:00 to 18:00
• Hybrid work format with office visits at least 3 days a week
• Comfortable modern office at: Moscow, Kutuzovsky Prospekt, 32
• Professional team (our center employs PhDs, graduate students, and Kaggle masters)
• Interesting tasks and active work with new technologies.