Seltz, a San Francisco-based startup, raised $12.5 million in seed funding to build a full-stack web search and knowledge infrastructure optimized for AI agents, LLMs, and RAG pipelines rather than human users.
Seltz.ai’s $12.5 million round was co-led by Speedinvest and B Capital, with participation from Italian Founders Fund, United Ventures, Future Back Ventures (Bain & Company’s venture arm), futurepresent, Arc Investors, Vento Ventures, Mango Capital, 2100 Ventures, and angels from companies including Google, Ramp, Tako, Hugging Face, and others.
What is Seltz.ai?
Seltz positions itself as the “Web Knowledge API” or “Web Knowledge Layer for AI Agents.” It provides fast, context-engineered, source-backed web data optimized for machine reasoning. Key differentiators include:
- Full-stack ownership: In-house crawler, index, retrieval, and ranking systems (hundreds of millions of pages crawled daily), avoiding reliance on Google, Bing, or other third-party APIs. This enables unique document coverage, reduced redundancy, and specialized vertical indexes.
- AI-native design: Returns full documents or precisely extracted passages/tables/images (not just snippets or links), with low latency (<200-250ms) and high token efficiency. Focus on “context engineering” — filtering/ranking for relevance to LLMs/agents.
- Enterprise features: Privacy-focused (end-to-end ownership, deployable on-premises/private cloud), SOC 2-aligned, high-throughput scalability, and domain filtering.
- Performance claims: Leads on Dynamic News Search Benchmark, 89% accuracy, ~5% ahead of next best; significantly faster than alternatives.
The company was founded in late 2025 (incorporated in the US) by CEO Antonio Mallia, who has a PhD in computer science from NYU focused on information retrieval. He previously worked as an applied scientist on Amazon’s AGI team and as a research scientist at Pinecone. The small team (~15 people, many with PhDs in IR, several ex-Amazon) operates remotely across the Bay Area, Pisa (Italy), and Leipzig (Germany).
Seltz already has an API (Python, Node.js, cURL clients), pilots with AI teams (including labs and agent builders), and testimonials from leaders at Tako, Synthesia, and Cohere emphasizing efficiency, data ownership, and production readiness.
Traditional search (e.g., Google) was built for short human keyword queries and link-skimming. AI agents issue long, precise, parallel queries at scale and need machine-readable, citable facts from page bodies/tables/etc., not click-bait snippets. Many existing AI search tools wrap legacy engines, inheriting their limitations in freshness, depth, control, and independence.
Seltz enters a hot but competitive space for “agentic” or AI-native search infrastructure. Notable peers include Parallel (recent $100M at $2B valuation), Exa ($85M), and Tavily (acquired for up to $400M). Google, OpenAI, Perplexity, Anthropic, and others are also investing heavily in their own crawling/indexing. Seltz bets that full-stack control, specialized architecture, and enterprise deployment options will win against wrappers and general-purpose giants.
The timing aligns with the rise of autonomous agents and RAG systems that require reliable, real-time external knowledge beyond training data cutoffs.
Proceeds will support:
- Scaling the search stack from initial verticals (e.g., strong news index) toward broad open-web coverage (tens of billions of documents).
- Team expansion (engineering/research focus).
- Initial enterprise sales and marketing efforts to reach more AI infrastructure teams.
Longer-term vision: An index shaped by agent interactions and feedback, creating a self-reinforcing system for AI-driven web exploration and knowledge assembly.
Recommended: Arcade Raises $60 Million In Series A Funding Round
Strengths and Strategic Positioning
- Technical moat potential: Deep expertise in IR, full ownership, and measurable early wins on benchmarks/speed.
- Founder-market fit: Mallia’s background directly addresses the shift from human to agentic search.
- Enterprise appeal: Data control, compliance, and deployment flexibility suit serious production users wary of third-party dependencies.
- Lean and capital-efficient start: Achieved product-market traction signals and a functional API with a tiny team in ~8 months.
Risks and challenges in this space include the extreme capital intensity of large-scale crawling/indexing, competition from better-resourced players (including hyperscalers and well-funded startups), legal/technical hurdles around web data access (e.g., anti-bot measures, lawsuits), and the need to rapidly scale sales/commercial execution while maintaining technical leadership.
The $12.5M seed validates strong investor conviction in the need for purpose-built web infrastructure for the agentic AI era. Seltz is positioned as a focused, technically ambitious contender aiming to become foundational for AI systems that must reason over live, dynamic web knowledge. Its success will hinge on execution in scaling the index, delivering consistent quality/latency advantages, and winning key enterprise and AI-lab customers.
Please email us your feedback and news tips at hello(at)dailycompanynews.com
