IT Press Tour 69: StratoFoundry

StratoFoundry joined the recent 69th edition of The IT Press Tour held last week in Ljubljana, Slovenia and it was the opportunity to discover a new player addressing data governance with advanced S3 management and Agentic AI.

The company has been founded in July 2026 by Emmanuel Forgues. He’s joined by Sami Fredj, CTO, Danice Cassam Chenaï, chief strategy officer, Thomas Raymond, chief AI officer, with a head of sales search underway as several POCs have been started.

Click to enlarge

The company is repositioning from an S3-based replication/resilience vendor into a “trusted data infrastructure” company. Its new platform, KARL, is pitched as a new infrastructure category, a Temporal Memory & Evidence Layer, not another repository, backup system, or AI memory tool. Forgues built several tools and services such as S3Guardian and S3Proxy now embedded into the new solution approach.

Click to enlarge

With the AI tsunami we all live in, the team has launched during the tour KARL for Knowledge, Analytics, Response & Learning. It addresses a classic challenge with object storage. in a nutshell, object storage scales content but loses the governed context (identity, policy, authority, approvals) behind past decisions. KARL demonstrated during a demo how current system state can no longer explain a past decision once policies/data change, KARL lets you query “what was true and valid at time T” with full evidence (data version, policy version, AI run, human approval).

Click to enlarge

Forgues insisted a lot on the architecture of KARL with a core component with several attributes and features: governed, versioned, never-rewritten object store built on S3-compatible storage, with a signed/chained journal leveraging Merkle trees coupled with RFC 3161 timestamps, content-addressed storage, indexing/graph layer, and verifiable “governed forgetting.” The product is written in Rust, multi-tenant by design, fail-closed access control (RBAC/ABAC), mTLS, per-tenant metering promoting a modern approach.

Click to enlarge

The different modules shown below explains the approach of KARL integrated with users’ environments and content servers such as file servers for isntance.

Click to enlarge

More globally, Forgues developed several independent services coupled with the various modules of KARL promoting a rich scalable but flexible model:

• S3Guardian – cross-endpoint S3/SaaS/on-prem replication (only product with published pricing: €9,900–€34,900+/year tiers, sold indirectly via certified partners)
• S3Proxy – stable “vBucket” abstraction for automatic redirection/parallel writes across S3 endpoints
• KARL Transit – verifiable, asynchronous portability of governed data/metadata/proofs via signed “KMLX” archives (zstd, SHA-256, Ed25519, optional AES-256-GCM); 14 Rust crates, 367 tests
• KARL Forge – “proves how a state became official” for AI-generated software: tracks requirement→spec→commit→test→verdict chain with a strict evidentiary hierarchy (cryptographically proven > rule-validated > tested > AI-evaluated/declared, which are non-evidentiary). 21 crates, 511 tests, 104 RPC methods
• KARL Foundry – sovereign local AI model factory: distills a remote teacher LLM (Kimi K3) into local 32B/70B specialist models on Nvidia GB10 hardware via a 12-gate (G0–G11) release pipeline, keeping raw client data local
• KARL Compass – AI-augmented “executive committee” (chief of staff-like AI) using paired local 30B/70B models plus remote-teacher escalation, covering strategy, legal/compliance, finance, etc
• KARL Audit, KARL Cyber, KARL IA Studio, KARL Media – continuous evidence capture, trusted-state incident recovery, AI/agent traceability, and governed multimodal search, respectively, mostly at prototype/controlled-demo stage

In terms of go-to-market, the team initiated a “Founding Design Partner Program”, up to 5 organizations, for paid pilots across Forge, Cyber and Audit tracks, no exclusivity, no product forking. Strategy is narrow prototyping first (deploy → demonstrate → measure → reference), with partner-channel scaling (MSPs, audit/compliance firms, AI integrators, DevSecOps specialists) under evaluation as a second phase. The roadmap has been also unveiled with a phase 1 to prove the core, a phase 2 to industrialize the knowledge layer, and a phase 3 to open the ecosystem via APIs/OEM. The team is very flexible to adopt any model and customize the product.

To conclude, it is important to understand that KARL reconstructs only captured/governed history, it doesn’t infer ungoverned data; “verifiable” means inspectable, not automatically true or certified; KARL complements and therefore doesn’t replace ERP, DMS, backup, and AI systems.