Python Developer (LLM / AI Infrastructure)

Quick Facts

• Role: Python Developer (LLM / AI Infrastructure)

Description

You will build infrastructure that manages what information enters an LLM’s context window. The work combines text processing, information retrieval, and agentic architecture, with a strong focus on measurable quality and cost using real benchmark evaluations.

Responsibilities

• Develop pipelines for text/context processing: segmentation, deduplication, scoring, and context-window selection

• Build and improve information retrieval components: indexing, ranking, embeddings, and search-quality evaluation

• Develop and test agent components: tool-use loops, memory/context management, orchestration

• Build and run benchmark evaluations to measure the impact of changes on quality and cost

• Prototype quickly and evolve prototypes into maintainable production systems

• Work in short iterations with direct feedback from a technical lead

Requirements

• Strong Python skills (data structures, complexity, profiling, clean testable code)

• Practical text-processing experience: tokenization, chunking, embeddings, and large-scale text handling

• Algorithmic thinking with ability to design and justify methods for segmentation/deduplication/scoring

• Comfortable working in Linux and using git

• Familiarity with tokenizers and how to apply them

• 1–2 years of commercial experience or very strong personal projects in LLM/NLP

• At least one of: LLM agent architecture (tool use, planning, context management), large-scale deduplication (e.g., MinHash/SimHash), working RAG/agent system used by others, open-source contributions in NLP/LLM tooling, or IR basics (BM25, vector search, precision/recall, NDCG)

Benefits

• Salary/B2B rate: 9,000–12,000 PLN on a contract (level depends on the technical interview)

• Work on AI infrastructure product with measurable results on public benchmarks

• Direct collaboration with a technical lead; fast growth, no intermediaries

• Flexible hours; office in Warsaw (Połczyńska 112A) with parking

• Access to paid AI tools (Claude, Copilot, Codex, ChatGPT, Gemini, etc.)

• Modern infrastructure work including B200/B300 clusters

• Training budget and participation in industry conferences

• Company hackathons