Senior AI Engineer – Legal RAG & Knowledge Systems

Level

Senior

Department

Other IT Positions

Type

Full Time

Project

NDA

Locations:

Ukraine

Europe

Remote

Job Details

Posted on:

January 22, 2026

Job ID:

About the Company

Established in 2004, ALLSTARSIT was founded with a clear vision: to enhance the landscape of global IT employment by bridging the gap between companies and skilled professionals. The core belief was that assembling a team shouldn't be hindered by geographical constraints. Fast forward to the present day, ALLSTARSIT stands as an international outstaffing service provider committed to change the way businesses recruit, compensate, and oversee top talent worldwide.

With operational hubs scattered across Europe, Asia, and LATAM, and its headquarters situated in San Francisco, US, the company boasts a workforce of over 1,000 adept professionals. Spanning across more than 20 countries, ALLSTARSIT offers a diverse range of skilled employees across various verticals, including AI, cybersecurity, healthcare, fintech, telecom, media, and so on.

About the Project

ALLSTARSIT is looking for a Senior LLM Engineer (RAG & Search, LangChain) for our client. They are building a core AI system that turns large collections of legal documents into an AI-queryable knowledge base — delivering reliable, grounded answers with precise citations to the source text.

This is a greenfield build: you’ll help shape the architecture, define quality standards, and take the system from early foundations to production-ready.

Specialization

Headquarters

Years on the market

Team size and structure

Current technology stack

Required skills:

Must-have

Strong Python
Hands-on experience building RAG / semantic search systems
LangChain (required): production experience with retrievers/chains/LCEL and orchestration of RAG pipelines
Experience working with unstructured documents (PDF/OCR/parsing/structuring) or deep RAG experience with solid ingestion pipelines
Strong engineering discipline: evaluation, experimentation, testing, and debugging
Working proficiency in English

Nice-to-have

OpenSearch / Elasticsearch (relevance tuning, BM25, hybrid search)
Vector search tooling (FAISS / pgvector / similar)
Reranking with cross-encoders (e.g., HuggingFace)
Experience with legal/regulatory text
AWS

Scope of work:

Own the end-to-end pipeline: legal documents → structured knowledge → retrieval → cited answers
Process and structure legal content from diverse formats (PDFs, scans, Word files, images):
- parsing / OCR / cleaning
- structural extraction (articles, sections, hierarchy)
- robust chunking with metadata and full traceability
Design and improve retrieval + RAG:
- hybrid search (BM25 + embeddings)
- reranking (cross-encoders)
- query understanding / expansion
Ensure grounded generation:
- strict source citations
- reliable no-answer / refusal behavior when evidence is missing
Measure and improve quality:
- build evaluation sets
- track retrieval metrics (Recall@k, MRR, nDCG)
- run regression tests and error analysis
Make it production-grade:
- stable pipelines, observability, performance tuning