Reach out directly about this role
🚀 Vacancy: Machine Learning Engineer / AI Engineer (Middle+)
💰 Rate: 250,000 RUB 📍 Location: Anywhere (remote) 📌 Format: Part-time possible We are looking for an engineer to build an AI core for an innovative B2C product on the React Native (Expo) + Supabase stack. You will be creating RAG systems and integrating LLMs into a universal cross-platform application.
🧠 Key tasks: • Development and optimization of RAG (Retrieval-Augmented Generation) systems. • Integration with LLMs (OpenAI, Anthropic, Google Gemini) via API. • Working with vector databases (pgvector/Supabase, Pinecone). • Development of AI logic on Edge Functions (Deno) and Python. • Collaboration with a hybrid team (management and domain — with the client).
🛠 Core stack: • AI: Python 3+, LangChain / Semantic Kernel, prompt engineering. • LLM: OpenAI, Anthropic, Google AI (Gemini). • Vector DBs: pgvector (in Supabase), Pinecone. • Backend: Supabase (BaaS — Auth, DB, Realtime, Storage), Edge Functions. • Frontend (for understanding): Expo (React Native) for Web/iOS/Android.
✅ Requirements (Must have): • 3+ years of experience with Python. • Practical experience with LLM APIs (OpenAI, Anthropic, Google AI). • Deep knowledge in RAG and working with vector databases (pgvector, Pinecone). • Experience using LangChain or Semantic Kernel. • Willingness to work with the stack: Supabase (BaaS), Expo, Edge Functions.
✅ Will be a plus (Nice to have): • Experience in prompt engineering and LLM quality evaluation. • Understanding or desire to learn Expo/React Native for Web. • Familiarity with DevOps: Vercel, EAS, GitHub Actions.
💎 What we offer: • Work on a modern product with a strong AI focus. • Remote format with the possibility of part-time work. • Rate of 250,000 rubles. • Flexible processes and work with experts in their field.
250,000 RUB
per month
from 3 years
Experience
Part-time
Employment
Remote
Work Format
Middle
Grade
Data Science & ML
Specialization
AI
Industry
By job title
Data Science & ML
Specialization
AI
Industry