I AM HIRED

Неделя 310 ₽Месяц 630 ₽3 месяца 1 680 ₽Год 5 680 ₽

AI Engineer, Retrieval (RAG) — Kosovo

ML и AI, AI-инженерия · вчера, 17:26 · 19.11.2026

Зарплата не указана

Откликнуться

Компания: ••••••••

Локация: Kosovo

Формат: не указан

Занятость: Полная занятость

Грейд: Senior

About MHRab5h7Y2

MHRab5h7Y2 is a workforce transition platform built for companies restructuring around AI. It gives founders, COOs, and Heads of People a live view of org structure, workforce costs, and scenario planning - so they can make workforce decisions with real data before committing to them.

Most companies are navigating some version of the same challenge: teams are changing, AI is being adopted, roles are shifting, and the cost of getting it wrong is higher than ever. But most leaders are making these decisions without seeing the full picture - what each role actually costs, where dependencies exist, and what could happen if someone leaves.

MHRab5h7Y2 brings org structure, workforce costs, and scenario modeling into one live view - with visibility into dependencies and knowledge risk when you need it. Not a static diagram someone updated last quarter. A living view of how your organization works and what a change could cost before you make it.

The Role

We are looking for a mid to senior AI Engineer to own retrieval quality end to end - from how documents are chunked and embedded to which candidates reach the model - and make Gm6M3EWVcH's answers measurably better and faster as the system scales to large, multitenant workloads. You will join Gm6M3EWVcH's existing engineering team of five and report directly to the Head of Development.

What You Will Own

• The ingestion pipeline: source-specific splitters and chunking, text-to-vector embedding (model in use Amazon Titan Embed v2, 1024 dimensions), embeddings stored in per-organization Postgres partitions using pgvector/pgvectorscale and DiskANN, and the hybrid keyword index (Postgres tsvector).

• The retrieval half of the query pipeline: dual vector search, hybrid merge and deduplication, candidate reranking, access control (RBAC) filtering applied inside the vector index, and scoped retrieval.

• Performance and scale: async and batched embedding calls, embedding throughput, DiskANN index health, and query latency versus recall as organizations grow.

• Evaluation: stand up the retrieval quality harness the system does not yet have — recall and precision at k, groundedness, and regression gates on any change to chunking, embeddings, index, or model.

Key Responsibilities

• Improve chunking fidelity (dates, attachments, authorship), embedding freshness, and hybrid search relevance.

• Ship date-aware retrieval - a document date column, range filters, ordering - and retrieval that supports streaming responses.

• Instrument retrieval quality and cost; drive decisions with evaluation data, not intuition.

What You Bring

Must Have

• Strong applied information retrieval and ML background: embeddings, vector search, hybrid retrieval, reranking, chunking strategy, and RAG evaluation methodology.

• Deep SQL and PostgreSQL skills (indexes, query plans). Hands-on experience with pgvector or another production vector index is a strong plus.

• Comfortable working in a PHP 8 and Laravel backend, or a strong general backend engineer willing to learn it. This is not a Python codebase.

• Experience with a managed inference layer (AWS Bedrock or equivalent).

Nice to Have

• DiskANN or HNSW internals; vector index tuning at scale.

Mindset

• Builder - energised by ambiguity and comfortable owning a system end to end.

• Rigorous - trusts evaluation data over intuition and builds the harness to prove it.

• Collaborative - works closely with a small, senior engineering team without ego.

• Curious about AI - genuinely excited by retrieval, search, and the knowledge intelligence space.

What We Offer

• Competitive salary and performance-based incentives.

• Fully remote-first environment with flexible working.

• A founding-stage role on a five-person engineering team, with direct influence over Gm6M3EWVcH's retrieval architecture.

• Fast-moving, international team with a strong culture of ownership.

• Learning & development budget.

Join our team at MHRab5h7Y2, a fast-growing global start-up. We're looking for driven, passionate individuals to join us as we continue to expand our reach and impact!

Поиск по вакансиям

Название компании скрыто

Названия компаний и их продуктов мы не публикуем на сайте — они доступны в Telegram-боте I AM HIRED. Откройте эту вакансию в боте: увидите название, сможете откликнуться и настроить рассылку подходящих вакансий.

Открыть вакансию в боте