UderTalk

Backend / AI Developer (Python, RAG, LangChain) at UnderTalk

About the project: Undertalk is an intelligent sales automation system. We are building AI agents LLM + RAG (Retrieval-Augmented Generation) that fully handle communication with clients in messengers: from the first consultation to closing the deal. The project is transitioning from an external team to internal development, so we are looking for a Problem Solver who will help build a reliable architecture and take responsibility for the product with a primary focus on building quality architecture and scaling AI solutions.

Key tasks:

• AI & RAG Development: Development and optimization of intelligent agents using LangChain. Implementation of complex interaction logic with OpenAI API (GPT-4, Embeddings, functional calls).

• Architecture & Data: Ensuring stable data storage and analytics in PostgreSQL. Building and maintaining a scalable database structure.

• Audit & Refactoring: Conducting a technical audit of the existing code, fixing bugs, and systematically migrating logic to a new three-tier architecture.

• Engineering Culture: Maintaining high development standards: writing tests (Unit/E2E), conducting Code Review, and maintaining technical documentation in Confluence.

• Autonomous Problem Solving: We work in short sprints (Jira) without micromanagement. We expect you to be able to independently identify bottlenecks in the system and propose technical solutions to eliminate them.

Technology stack and tools:

We are looking for a developer who not only knows Python but can build systemic solutions using modern AI-native approaches.

1. Core Backend & Architecture:

• Framework: FastAPI (asyncio, asynchronous development).
• Database & ORM: PostgreSQL + SQLAlchemy (model design, working with migrations).
• Architecture: Mandatory use of three-tier architecture (Routes-Services-Repositories). Clear separation of business logic and data access layer.
• Validation: Pydantic v2 for validation and data schemas.

2. AI & RAG Specialized:

• Framework: LangChain (our main tool for building agent logic).
• LLM: Deep integration with OpenAI API (GPT-4, Embeddings).
• RAG: Understanding of Retrieval-Augmented Generation mechanisms (working with context, search, reranking).

3. AI-Driven Development (Productivity Stack):

• IDE: Experience with Cursor (using AI features to speed up writing and refactoring code).
• Cloud: Experience with CloudCode (integration with cloud infrastructure).
• Efficiency: Ability to use AI tools for automating test writing and documentation.

4. Infrastructure:

• Containerization: Docker and Docker-compose (preparing development and deployment environments).
• CI/CD: Working with Git (GitHub/GitLab), understanding code delivery processes.

Conditions:

• Remote work with a flexible schedule.
• No unnecessary bureaucracy and the ability to influence the tech stack.
• No micromanagement.

To apply:

• Link to GitHub (preferably with code examples on FastAPI or LangChain).

• Current resume.

• Brief description of your experience with RAG: what tools you used, which databases you worked with, and how you addressed response quality issues (e.g., working with context or prompts).

• Salary expectations (USD/hour).

The selection process includes a short technical task (up to 2 hours) to check the code writing culture.