Omer Primor

Omer Primor leads product marketing at Bright Data, where he focuses on turning the constantly changing web into dependable information for AI agents. His work on Context-as-a-Service addresses a practical architectural question: when should organizations purchase ready-made context, and when should they build infrastructure to collect, own, and refresh it themselves?

From maritime intelligence to agent infrastructure

Primor studied history and sociology at Tel Aviv University before joining Windward, a maritime intelligence company, where he worked in product management and business development and researched manipulation of vessel-tracking data. He subsequently became Windward’s head of product marketing and then head of marketing.

His writing on human-machine collaboration advocated automating data preparation while preserving analysts’ judgment. Another analysis of persistent maritime monitoring examined how historical behavior, global vessel tracking, and complementary datasets produce stronger operational intelligence. His investigation of the Grace 1 tanker illustrated how suspicious movements and potential sanctions evasion become visible through patterns over time.

Primor later served as vice president of marketing at Imvision, an API-security company acquired by Intuit. His writing on security-team collaboration explored how organizations can protect APIs throughout development and production. At Bright Data, he shifted that experience with operational intelligence and enterprise infrastructure toward the web-data requirements of agentic systems.

The economics of useful context

In his AI Engineer World’s Fair session, Primor develops several concrete positions:

• Context-as-a-Service: Specialized providers collect, deduplicate, enrich, and structure web information into entities and relationships that agents can access through APIs or the Model Context Protocol.
• Longitudinal context: Point-in-time search can identify a company’s current vacancies or a product’s latest price, but persistent research requires historical records and continual updates to understand how those facts change.
• Retrieval frequency drives cost: Repeated requests can incur provider fees and model-processing expenses even when agents revisit substantially the same information.
• Owned versus rented context: An informal company-enrichment experiment compared search tools, commercial context providers, and a quickly assembled in-house pipeline. Primor found that directly collected data may become more economical for recurring workloads, depending on setup costs, coverage requirements, and query volume.
• Web context engineering: Exploratory questions, narrowly defined domains, and persistent research warrant different retrieval strategies; search services, specialist providers, and owned pipelines each suit different operational needs.