Senior ML Backend Engineer - Search & Recommendations

Salla; Makkah Region; On-site; SAR 300,000 - 600,000; Full time

Job summary

Salla is the leading e-commerce platform in MENA, powering a discovery stack for search and recommendations. You will own the platform supporting fast, fresh and reliable ranking for shoppers, collaborating with data scientists and ML engineers on scalable backends.

Join the ML Platform team to design, build, and operate low-latency services, streaming ingestion, and integration with product surfaces. This is a hands-on role with ownership of deployment and experimentation.

Qualifications

• 5+ years building backend or platform systems in production.

Responsibilities

• Build and run query and recommendation serving: retrieval, score fusion, filtering, caching.
• Hold tight latency budgets, including double-digit-millisecond p95 for search-as-you-type.
• Design graceful degradation, so serving holds up when a model or vector path is slow or down.
• Build experiment plumbing: shadow traffic, A/B assignment, guardrail metrics, rollback.
• Build change-data-capture from source systems into search indexes and feature stores, with measured freshness.
• Keep embeddings fresh for new and updated items.
• Run zero-downtime reindexing of multi-million-document collections.
• Build feature logging that turns production traffic into training data for ranking models.
• Ship changes into product teams' backends and clients: new endpoints, cutovers, experiment flags.
• Work with product engineering teams on reviews, release windows and production ownership.
• Monitoring, alerting, runbooks, and your share of on-call for what you ship.

Skills

Backend systems

Production search engine

Low latency

Streaming ingestion

Python

TypeScript

A/B testing

Claude Code

Tools

Elasticsearch/OpenSearch/Solr/Vespa

Kafka/Debezium

Job description

You'll own the online systems that take a catalogue change to a ranked result in front of a shopper: search, retrieval and recommendation serving, the streaming data that keeps them fresh, and the experiments that ship them.

Salla is the leading e-commerce platform in MENA, serving about 68,000+ merchants. Our AI team is building a self-owned discovery stack across our marketplace and merchant stores: hybrid lexical and vector search, Arabic-first query understanding, and learned ranking shared between search and recommendations. Data scientists own the relevance and ranking models. You own the platform that makes them fast, fresh, reliable and shippable.

The work is roughly half online serving, a third streaming data, and the rest integration into product surfaces. You sit on the ML Platform team spending time between search and serving for our recommendation systems and work day to day with search data scientists and the ranking lead

What you'll do

Online serving

• Build and run query and recommendation serving: retrieval, score fusion, filtering, caching.
• Hold tight latency budgets, including double-digit-millisecond p95 for search-as-you-type.
• Design graceful degradation, so serving holds up when a model or vector path is slow or down.
• Build experiment plumbing: shadow traffic, A/B assignment, guardrail metrics, rollback.

Streaming data

• Build change-data-capture from source systems into search indexes and feature stores, with measured freshness.
• Keep embeddings fresh for new and updated items.
• Run zero-downtime reindexing of multi-million-document collections.
• Build feature logging that turns production traffic into training data for ranking models.

Integration

• Ship changes into product teams' backends and clients: new endpoints, cutovers, experiment flags.
• Work with product engineering teams on reviews, release windows and production ownership.

Operability, throughout

• Monitoring, alerting, runbooks, and your share of on-call for what you ship.

• 5+ years building backend or platform systems in production.
• 2+ years operating a production search engine (Elasticsearch, OpenSearch, Solr, Typesense, Vespa or similar): index lifecycle, analyzers, reindexing at scale, debugging relevance issues with data scientists.
• Low-latency services under real traffic, and the ability to explain where the p95 goes.
• Streaming ingestion: Kafka, Debezium or similar change-data-capture, plus idempotent backfills.
• Strong Python. Comfortable shipping in TypeScript.
• Shipped changes through A/B tests and staged rollouts.
• Daily use of agentic coding tools such as Claude Code. We expect one engineer plus agents to cover what used to take a team.