Infino AI
Open-source retrieval engine for AI agents, combining search, vector embeddings, and SQL on object storage.
Website: https://infino.ai/
Cover Block
Public sources
| Field | Value |
|---|---|
| Name | Infino AI |
| Tagline | Open-source retrieval engine for AI agents, combining search, vector embeddings, and SQL on object storage [Infino AI, about] |
| Headquarters | Rehoboth Beach, DE |
| Stage | Seed [Ground.news, October 2026] |
| Business model | Open Source / Commercial |
| Industry | Deeptech |
| Technology | AI / Machine Learning |
| Growth profile | Venture Scale |
| Founding team | Co-Founders (4): Ekechi Nwokah, Vinay Kakade, Asif Makhani, Murali Krishna [Infino AI, about] |
| Funding label | Seed |
| Total disclosed | ~$7,500,000 [Ground.news, October 2026] |
Links
Public sources
- Website: https://infino.ai/
- LinkedIn: https://www.linkedin.com/company/infino-ai
Executive Summary
PUBLIC Infino AI is building an open-source retrieval engine for AI agents, and it merits attention now because the company has paired a technically ambitious product claim with a newly reported $7.5 million seed round led by Bessemer Venture Partners [Infino AI, about] [Ground.news, October 2026]. The company has emerged from stealth around a simple thesis: developers and data teams should be able to run SQL, full-text search, and vector retrieval over a single copy of data in Apache Parquet on object storage, rather than maintain separate search clusters and vector databases [Infino AI, about] [GitHub].
The public founding record is still thin, but Infino names Ekechi Nwokah, Vinay Kakade, Asif Makhani, and Murali Krishna among the people building the company, with third-party and profile-based sources linking those founders to prior work in search, cloud infrastructure, and applied AI [Infino AI, about] [Chandra Sekar Srinivasan - Amazon | LinkedIn] [Prakash Nagarajan - Twilio Inc. | LinkedIn]. That background is the main reason the story is credible at this stage: Kakade is reported to have led development tied to AWS Elasticsearch Service and previously co-founded FinitePaths, which Lyft acquired, while Makhani previously served as CTO of Handshake and worked on Amazon CloudSearch, LinkedIn Learning, and A9 according to public profiles and company databases [Business Standard, March 2017] [FundraisingFox] [PR Newswire, January 2021].
On product, the differentiation rests on collapsing retrieval and query workloads into the storage layer itself. Infino says the engine combines BM25-style full-text search, vector embeddings, and SQL directly on Parquet files stored in Amazon S3, Google Cloud Storage, or Azure Blob, and its GitHub description similarly says the system runs SQL, full-text search, and vector search over one copy of data on object storage [Infino AI, about] [GitHub].
The commercial model appears to be open source plus paid production usage, with pricing tied to a write token and production billing metering the work performed by writes [Infino AI, pricing]. That model can fit infrastructure adoption patterns, but the evidence base does not yet verify customers, revenue, or production deployments, so the near-term underwriting case still turns more on technical adoption than on commercial proof [FundraisingFox] [Infino AI, pricing].
Over the next 12 to 18 months, the key questions are whether Infino can convert architectural interest into repeatable production usage, whether benchmark claims on ingest and retrieval hold up outside company-authored materials, and whether the team can define a durable wedge against incumbent search, vector, and lakehouse stacks [Infino AI, blog] [The New Stack] [Dealroom News]. For investors, the setup is promising but early: strong founder-market fit is the visible asset, while customer validation remains the main gap in the public record [Infino AI, about] [Ground.news, October 2026].
Lightly corroborated -- This section relies on one independent funding report, one independent founder background source set, and multiple company-authored product claims with limited third-party corroboration.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | Open Source / Commercial |
| Industry / Vertical | Deeptech |
| Technology Type | AI / Machine Learning |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding | Seed, total disclosed about $7.5 million |
How the Company Got Here
PUBLIC
Infino AI is presenting itself as infrastructure for a specific bottleneck in agentic software: retrieval across data that already lives in object storage. On its website, the company says it is building an open-source retrieval engine that combines search, vector embeddings, and SQL over Apache Parquet files stored on services such as Amazon S3, Google Cloud Storage, and Azure Blob [Infino AI]. The same public materials place the company in Rehoboth Beach, Delaware, and describe its mission in short form as “Building retrieval for agents” [Infino AI].
The public record is thin on formal company history, and that matters for chronology. Infino’s site names Ekechi Nwokah, Vinay Kakade, Asif Makhani, and Murali Krishna as the team behind the business, but the available company materials used here do not provide a founding date or legal entity details beyond the Delaware headquarters reference [Infino AI]. A Crunchbase profile for Ekechi Nwokah identifies him separately as co-founder and CEO of Migo and previously a principal security engineer at A9.com, which helps establish prior operating background but does not fill in Infino’s incorporation timeline [Crunchbase].
The clearest public milestone so far is the company’s emergence from stealth alongside a disclosed $7.5 million seed round in September 2026, led by Bessemer Venture Partners, according to Ground.news news, October 2026]. Infino’s homepage also states that it “raises $7.5M from Bessemer to build one retrieval layer for all your agents,” which aligns with the funding announcement but remains company-published confirmation rather than an independent filing [Infino AI]. On the evidence available for this section, the company is early, newly public, and still lightly documented outside its own web presence and secondary database profiles [Infino AI] [Crunchbase].
Lightly corroborated -- This section relies primarily on company website disclosures, with partial corroboration from Crunchbase and one secondary funding report.
Product and Technology
Sources and analysis
Infino is making a fairly specific technical claim, and the interesting part is where the system sits in the stack rather than any one model feature. The company says its product is an open-source retrieval engine that combines full-text search, vector embeddings, and SQL directly on Apache Parquet files stored in object storage such as Amazon S3, Google Cloud Storage, and Azure Blob [Infino AI, about]. Its homepage frames the pitch as "scalable search + analytics infrastructure on object storage, for 10x cheaper," while GitHub materials describe the engine as running SQL, BM25 full-text search, and vector search over a single copy of data on object storage [Infino AI, homepage] [GitHub].
The public positioning suggests Infino is trying to collapse parts of the modern retrieval stack into one layer. According to the company, that means developers building retrieval-augmented systems and AI agents may not need a separate search cluster and vector database, while data teams can query warehouse-style data with the same substrate [Infino AI, about]. The site also says agents can express complex questions in a single SQL query, and product materials describe hybrid retrieval that "retrieves and fuses candidates for hybrid search" [Infino AI, homepage] [Infino AI, agents].
A second thread in the product story is operational control, although most of that evidence remains company-supplied. Infino says billing is tied to ingest work through a write-token model, with production metering based on the work each write performs [Infino AI, pricing]. On security and deployment, the company says object-store access uses ambient cloud identities and that operational secrets come from configuration and cloud secret-management systems, and a company blog post describes the product as a governed execution layer for AI agents with audit trails, lineage, and centralized RBAC [Infino AI, security] [Infino AI, blog]. The only public performance figure in the available material is a company-posted benchmark of about 33,100 documents per second for full-text ingest and 62,200 documents per second for vector ingest at dimension 384 on a 10 million document table on S3, under a specified 8-core test configuration; that figure is useful as a product signal, but it is not independently verified [Infino AI, blog].
Lightly corroborated -- Product architecture is partially corroborated by company materials and GitHub, but performance, cost, and governance claims are primarily company-reported.
Where the Demand Sits
PUBLIC
The market matters now because retrieval has moved from a supporting data function to a core control point for enterprise AI systems, yet the public evidence on Infino's exact served market remains thin.
No cited third-party market report in the provided record establishes a direct TAM, SAM, or SOM for open-source retrieval engines on object storage. The closest public framing is analogous rather than exact: Infino positions itself as infrastructure that combines SQL, full-text search, and vector search over a single copy of data in Apache Parquet on object storage, aimed at developers building AI agents and data teams querying warehouse-scale data [Infino AI, about] [GitHub]. That places the company at the intersection of several larger software budgets, including enterprise search, vector databases, data lakehouse query engines, and emerging agent infrastructure, but the available sources do not quantify how much of those adjacent budgets is realistically contestable by Infino today [FundraisingFox] [The New Stack].
The immediate demand signal is architectural fragmentation. Infino's core claim is that developers otherwise piece together separate search clusters, vector databases, and analytics systems to support retrieval-augmented generation and agent workflows, and that its engine can collapse that stack onto object storage such as Amazon S3, Google Cloud Storage, or Azure Blob [Infino AI, about] [GitHub]. The public sources also show why that pitch may resonate now: the company emerged from stealth in 2026 with a $7.5 million seed round led by Bessemer Venture Partners, suggesting investors see timing in the convergence of agent tooling and lower-cost data infrastructure [Ground.news, October 2026] [Dealroom News].
table
| Market frame | What the cited evidence supports | Source basis |
|---|---|---|
| Core served market | Retrieval infrastructure for AI agents over object storage | [Infino AI, about] [GitHub] |
| Adjacent market 1 | Enterprise and developer search workloads using BM25 or full-text retrieval | [GitHub] [Infino AI, homepage] |
| Adjacent market 2 | Vector retrieval for embedding-based AI applications | [GitHub] [Infino AI, about] |
| Adjacent market 3 | SQL analytics on Parquet and warehouse or lake data | [Infino AI, about] [Infino AI, homepage] |
| Substitute budget | Separate search clusters and vector databases | [Infino AI, about] [The New Stack] |
The table points to a market that is less a single clean category than a budget-consolidation wedge. If Infino works as described, its opportunity depends on taking spend from multiple incumbent layers rather than creating a standalone line item from scratch [Infino AI, about] [The New Stack].
Adjacent and substitute markets matter because buyers may compare Infino less against one direct peer than against existing combinations of tools. A data platform team already using object storage, Parquet, and cloud IAM may view Infino as an extension of lakehouse analytics into AI retrieval, while an application team may compare it against a dedicated vector database plus Elasticsearch or OpenSearch-style search infrastructure [Infino AI, security] [GitHub]. That broadens theoretical demand, but it also means adoption may hinge on where budgets sit inside the organization, whether platform teams own agent infrastructure, and whether engineering leaders prefer consolidation over specialist systems. The current public record does not show which buyer persona is converting first.
Regulatory and macro forces appear supportive in principle, though the evidence here is mostly product-adjacent rather than market-wide. Infino's public materials emphasize audit trails, immutable lineage, centralized RBAC, cloud secret management, and ambient cloud identities, all features that map to enterprise requirements in sensitive environments [Infino AI, blog] [Infino AI, security]. In practical terms, that suggests the company is aligning its product with a period when enterprises want AI systems to run closer to governed data estates rather than copy data into separate services. The limitation is that no independent source in the provided materials verifies customer demand from regulated sectors, procurement timelines, or budget conversion rates.
Company-stated, unverified -- This section relies primarily on company materials and adjacent-source interpretation, with no cited third-party market sizing report for the exact category and only partial independent corroboration from GitHub and secondary coverage.
Competitive Landscape
MIXED Infino is positioning itself less as another database and more as a retrieval layer that tries to collapse search, vector retrieval, and SQL onto a single Parquet copy in object storage, which places it at the intersection of search infrastructure, vector databases, and lakehouse-adjacent query engines [Infino AI, about] [GitHub] [The New Stack].
The competitive map is easier to understand by splitting the market into substitutes rather than looking for one direct peer. In full-text and relevance search, the obvious incumbent reference point is Elastic, which Infino itself signals by using language such as "replace elastic" on its homepage [infino.ai]. In vector retrieval and hybrid search, the substitute set includes purpose-built vector databases and retrieval stacks, although the structured research here does not verify named vendors beyond Infino's own framing, so the safer reading is that Infino competes against a category of separate search clusters plus vector databases rather than a single company [Infino AI, homepage] [Infino AI, about].
A second flank comes from adjacent data infrastructure. Infino's claim is that SQL, BM25, and vector search can run over one copy of data on object storage, which means the practical alternative for some teams is not a search engine at all but the existing warehouse or lake stack they already operate [GitHub] [Infino AI, about]. For developers building retrieval-augmented systems, the choice may be between adopting a dedicated retrieval layer like Infino or stitching together warehouse queries, embedding pipelines, and a separate index. For data teams, the substitute is often internal tolerance for that fragmentation rather than a formal vendor bake-off.
Where the company appears strongest today is founder-market fit. Public materials tie the team to Amazon search infrastructure, AWS OpenSearch, Handshake, Lyft, LinkedIn, and A9, with Vinay Kakade's prior company FinitePaths reported as acquired by Lyft and Asif Makhani reported as a former Handshake CTO [FundraisingFox] [Business Standard, March 2017] [PR Newswire, January 2021] [Bloomberg Markets]. For an infrastructure startup selling a technical architectural bet, that background matters because early adoption often tracks credibility with engineers before it shows up as broad distribution. The complication is durability: talent is a real edge in the first product cycles, but by itself it is perishable unless it converts into ecosystem pull, production references, or a technical moat that others cannot reproduce.
The most exposed point is go-to-market. The public record here does not verify named customers, production deployments, or partnerships, while the core product message asks buyers to replace or avoid multiple entrenched systems at once [FundraisingFox] [Infino AI, about]. That is a large architectural ask. Elastic retains the advantage of an established installed base if the workload is primarily search, and warehouse-centric approaches retain the advantage of already being budgeted and governed if the workload is primarily analytics. Infino's proposition is strongest only if hybrid agent retrieval becomes important enough that teams feel the pain of running separate search, vector, and SQL systems on the same data.
The most plausible 18-month scenario is that the category fragments by buyer urgency rather than converging around one architecture. Elastic looks like the winner if enterprises continue to prioritize incumbent reliability and incremental extension of existing search estates, because it already owns the search budget and operational muscle in many organizations [infino.ai]. Infino is the winner if agent builders increasingly insist on one retrieval layer over object storage and are willing to trade incumbent familiarity for a simpler data path, especially in teams that already treat Parquet on S3, GCS, or Azure Blob as the system of record [Infino AI, about] [GitHub]. The loser if that shift does not happen is any newcomer, including Infino, whose differentiation rests on unifying categories that some buyers may still prefer to source separately.
Lightly corroborated -- Competitive positioning is supported by Infino's public product materials and GitHub, but named competitor evidence in the source set is thin and the section relies in part on category inference rather than independently reported market-share or customer-switching data.
Opportunity
PUBLIC
If Infino executes, the prize is not a point solution in retrieval, it is becoming a default data access layer for AI agents that need search, vector retrieval, and SQL over the same object-store data plane [Infino AI] [GitHub] [Ground.news, October 2026].
The headline opportunity is straightforward: collapse a fragmented retrieval stack into one system that developers can adopt inside the storage layer they already use. The public evidence is still thin, but the core product claim is specific enough to matter. Infino says it runs SQL, BM25 full-text search, and vector search over a single copy of data stored on object storage, including Amazon S3, Google Cloud Storage, and Azure Blob [Infino AI] [GitHub]. If that architecture holds up in production, it addresses a real operational pain point for teams that otherwise maintain separate warehouses, search clusters, and vector databases for agent and RAG workloads [Infino AI] [Dealroom News].
The reason that upside is reachable, rather than pure aspiration, is the mix of product positioning and team history. The company emerged from stealth with a reported $7.5 million seed led by Bessemer Venture Partners, which at minimum suggests institutional interest in the retrieval-layer thesis [Ground.news, October 2026]. The founding bench also maps closely to the problem: public profiles and company materials tie Vinay Kakade, Asif Makhani, Ekechi Nwokah, and Murali Krishna to prior work in search, distributed systems, AWS search infrastructure, A9, LinkedIn, and Handshake [Infino AI] [Business Standard, March 2017] [PR Newswire, January 2021] [Crunchbase]. In infra, that background does not guarantee distribution, but it does improve the odds that the technical wedge is real.
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Open-source control plane for agent retrieval | Infino becomes the preferred retrieval engine for teams building agents on data lakes and object storage, first through developer adoption and later through commercial governance and billing layers | A widely adopted open-source release paired with production features such as governed execution, audit trails, lineage, and centralized RBAC [Infino AI, blog] [GitHub] | The company already presents itself as open source with a commercial motion layered on top, and its messaging is tightly aligned with agent builders who want one retrieval layer instead of several systems [Infino AI] [Dealroom News] |
| Displace parts of the search plus vector stack in cloud data estates | Infino wins where enterprises want retrieval on existing Parquet data without standing up separate search clusters or vector databases | Proof that one-copy retrieval on object storage is materially cheaper or simpler than the incumbent architecture, which Infino frames as "10x cheaper" on its homepage [Infino AI] | The product is explicitly built around Apache Parquet on object storage and claims to eliminate separate search clusters and vector databases, which matches a visible enterprise cost-control theme in AI infrastructure [Infino AI] [The New Stack] |
| Become the governed execution layer for sensitive agent workloads | Infino moves from retrieval engine to policy and execution layer for regulated or security-sensitive deployments | Demand for auditable agent workflows in enterprise settings, especially where data access control matters as much as model quality [Infino AI, security] [Infino AI, blog] | The company already describes a governed execution layer with immutable lineage, full audit trails, centralized RBAC, ambient cloud identities, and cloud-secret integration. Those are enterprise buying signals if they are mature in practice [Infino AI, blog] [Infino AI, security] |
What compounding looks like here is not a consumer network effect. It is architectural entrenchment. If developers begin with Infino for open-source retrieval on Parquet, then add production billing, governed execution, and security controls as workloads move into sensitive environments, the company can turn a low-friction adoption wedge into a harder-to-replace control point [Infino AI, pricing] [Infino AI, security] [Infino AI, blog]. The early signs are directional rather than conclusive: the product already spans ingest metering, production billing based on write work performed, hybrid retrieval, and identity-aware object-store access, which implies the company is designing for an operating layer rather than a narrow query feature [Infino AI, pricing] [Infino AI, agents] [Infino AI, security].
There is also a technical compounding story if the performance claims stand up under independent use. Infino has published a benchmark claiming about 33,100 documents per second for full-text ingest and 62,200 documents per second for vector ingest at dimension 384 on a 10 million-document table on S3, using 8 cores, 256 superfiles, and 16 commits [Infino AI, blog]. That is company-published evidence, so it should be treated cautiously. Still, in infrastructure markets, a credible performance narrative can matter well before revenue is visible, because developer trust and architectural standardization often begin with benchmark-backed experimentation.
The size of the win is easiest to frame through strategic position rather than reported market size, because no credible TAM figure is present in the public evidence. If Infino becomes an adopted retrieval and governance layer for agent workloads across lakehouse data, the closest upside pattern is a core infrastructure company that sits on the query path for high-value applications. With only a seed round publicly reported and no revenue disclosed, any valuation framing here is necessarily a scenario, not a forecast. Under the strongest scenario, where Infino becomes a standard retrieval substrate for enterprise agent systems, the company could plausibly support venture-scale outcomes in the multi-billion-dollar range (scenario, not a forecast), because control points in data infrastructure that combine developer adoption with enterprise governance have historically captured outsized value relative to single-feature tools [Ground.news, October 2026] [The New Stack] [Dealroom News]. Public evidence does not yet justify a tighter valuation range than that.
Lightly corroborated -- Material claims rely on Infino's own website and blog, with partial corroboration from GitHub, Ground.news, Business Standard, PR Newswire, and Crunchbase.
Sources
Public sources
[Ground.news, October 2026] PERPLEXITY SONAR PRO BRIEF | https://ground.news/article/pakistani-origin-founders-startup-infino-ai-raises-75m-seed
[TechCrunch, August 2018] Offering a white-labeled lending service in emerging markets, Mines raises $13 million | https://techcrunch.com/2018/08/10/offering-a-white-labeled-lending-service-in-emerging-markets-mines-raises-13-million/
[Business Standard, March 2017] Lyft acquires social Q&A platform FinitePaths | https://www.business-standard.com/article/companies/lyft-acquires-social-q-a-platform-finitepaths-117030200323_1.html
[PR Newswire, January 2021] Handshake Appoints Former Amazon and Lyft Product and Technology Leader Asif Makhani as Chief Technology Officer | https://www.prnewswire.com/news-releases/handshake-appoints-former-amazon-and-lyft-product-and-technology-leader-asif-makhani-as-chief-technology-officer-301201918.html
[infino.ai] Infino | Answer any agent question on Parquet. | https://infino.ai/
[infino.ai] PERPLEXITY SONAR PRO BRIEF | https://infino.ai/about/
[linkedin.com] Infino AI | LinkedIn | https://www.linkedin.com/company/infino-ai
[tracxn.com] Infino - 2026 Company Profile, Team, Funding, Competitors & Financials - Tracxn | https://tracxn.com/d/companies/infino/__EtoGjbLX7VFWz9T12lD71t26JYoEfTUZciIdWylFMe0
[fundraisingfox.com] PERPLEXITY SONAR PRO BRIEF | https://fundraisingfox.com/companies/infino
[TechCrunch, December 2019] Credit startup Migo expands to Brazil on $20M raise and Africa growth | https://techcrunch.com/2019/12/03/credit-startup-migo-expands-to-brazil-on-20m-raise-and-africa-growth/
[TechCrunch, February 2023] AI is the next frontier , but for whom? | https://techcrunch.com/2023/02/08/ai-is-the-next-frontier-but-for-whom/
[TechCrunch, October 2018] Africa Roundup: Paga goes global and 4 startups raise $99M in VC | https://techcrunch.com/2018/10/03/africa-roundup-paga-goes-global-and-4-startups-raise-99m-in-vc/
[TechCrunch, May 2019] These startups are locating in SF and Africa to win in global fintech | https://techcrunch.com/2019/05/22/startups-locating-in-sf-and-africa-to-win-global-fintech/
Articles about Infino AI
- Infino AI Emerges From Stealth With a Rust-Based Retrieval Engine — The startup, founded by veterans of Amazon and LinkedIn search teams, aims to simplify the data stack for developers building AI agents.