Backend Engineer (Python / Social Media Parsing / LLM)

We collect and classify trends in social media niches. The system parses platform content, labels it by topic and format using LLM, and provides the result in expert content plans.

Currently, the module runs on third-party paid services, is launched manually, and takes several days to analyze a niche due to download limits. The engineer's task is to migrate it to custom parsers and automate its launch, expand the list of platforms, and reduce the cost of execution to near zero.

The project is in the MVP stage. The first experts are expected to start using the product in August 2026, with the first clients anticipated around September. The backlog for the next 6-12 months is still being formed, so tasks and priorities will be clarified as the product is tested.

Responsibilities

1. Write and maintain social media parsers: Instagram*, TikTok, Threads, followed by Facebook*, X, and other platforms.
2. Transition from third-party paid services to Apify and then to custom solutions, controlling execution costs.
3. Implement automatic pipeline launch: schedules, queues, retries, monitoring, and failure alerts.
4. Classify content using LLM: topics, formats, trend indicators.
5. Build and manage processing pipelines using LangGraph or LangChain.
6. Compare models by quality and cost, and switch to more cost-effective labeling.
7. Design data storage and data marts for content plan generation.
8. Maintain system stability amidst changes on the platforms' side.

Tasks are not provided as ready-made technical specifications but as descriptions of the process and expected results. The engineer needs to independently clarify limitations, compare implementation options, and choose a solution suitable for the current scale, not a hypothetical one.

Team

The engineer will work directly with the Product Manager and full-stack engineers of the platform that consumes trend-watching data. There is no CTO, dedicated architect, or DevOps in the team yet, so the employee must make technical decisions independently and bring tasks to production.

Expansion of the technical team is planned for later.

Stack

Backend: Python, FastAPI, asynchronous processes, queues, schedulers, external APIs. Parsing: custom parsers, Apify, headless browsers, working with proxies and limits. LLM: Gemini API, Anthropic, and comparable providers, LangGraph, LangChain, structured output.

Data and infrastructure: PostgreSQL, S3-compatible storage, Railway, Git.

The stack is negotiable: the engineer can suggest other tools if they better suit the product's tasks.

What's Important

1. Five years of commercial development experience and working at a senior level.
2. Strong Python skills and experience launching production applications.
3. Industrial experience in social media parsing: pagination, limits, proxies, resilience to platform-side changes.
4. LangGraph or LangChain experience is mandatory, with production experience, not just in prototypes.
5. Practical experience with Gemini API and other models, ability to compare them by quality and cost.
6. Content classification and labeling through LLM, structured output, quality control of results.
7. Experience with queues, schedules, retries, and idempotency for long-running processes.
8. Solid knowledge of PostgreSQL and schema design for analytical data.
9. Cost control: understanding the components of execution cost and the ability to reduce it.
10. Understanding of legal limitations and platform rules, willingness to discuss risks before implementation.

AI Tools

In daily development, we use Claude Code for working with the codebase, testing hypotheses, and accelerating implementation.

The engineer must understand and verify generated code, identify errors, and be responsible for the quality of the final solution. Experience with Claude Code will be an advantage, but systematic work with Cursor, GitHub Copilot, Windsurf, or other AI tools will also suffice.

Will be a plus

— Apify, Playwright, Puppeteer — LLM system quality assessment: test sets