ANDRIY BAYGERYCH
Founder-Operator | Applied AI Systems Engineer
Norway | info@infinity-folder.no | enterprise@infinity-folder.no | +47 904 16 707
Portfolio: https://infinity-folder.no
Live substrate: https://infinity-folder.org/index.html
Public record: https://github.com/Nordvei/sage-track-record

PROFILE
Independent founder-operator of IFP/SAGE, a local-first governed evidence and learning substrate. I work across live-system re-derivation, agent infrastructure, independent verification, authority containment, machine-readable evidence, and provider-neutral worker steering. I coordinate high-compute AI workers while keeping accepted state, verification, closure, and authority host-owned.

NON-TRADITIONAL PATH
No formal IT engineering diploma is claimed. Capability is presented through live systems, exact artifacts, bounded case studies, and visible negative results.

CORE CAPABILITIES
- Applied AI systems architecture
- Live-system and repository re-derivation
- Writer-reader-consumer and authority-path analysis
- Agent evals, closure, and L0-L4 learning classification
- Provider-neutral WorkerDoor and capability-envelope design
- Independent verification and exact-artifact qualification
- Admission boundaries and authority containment
- Append-only evidence, cryptographic receipts, and typed correction
- Human and machine-readable public truth surfaces
- Multi-model technical programme direction

SELECTED WORK
1. SAGE Machine Governor P0-RL (August 2026, provisional)
   Mapped 35 learning/control mechanisms; 21 had current firing evidence or output; zero verified L2, L3, or L4 external-worker loops; one thin deterministic steering consumer identified. Formal artifact review remains pending.

2. DecisionBundleV1 P1 (August 2026, qualified/bounded)
   315 passing clean-worktree tests. Package internal consistency verified. Broader full_record remains UNESTABLISHED.

3. OVL paid attestation rail (June 2026, closed on testnet)
   Base Sepolia x402 settlement, Ed25519 signature, independent content-hash re-derivation, ledger correlation, and 4/4 tamper rejection. Self-paid mechanism proof, not customer demand.

4. Energy prediction truth correction (August 2026, falsified/held)
   Governed result approximately 46.5 percent with no demonstrated advantage over persistence. Previous 65 percent headline superseded. Public timestamped record retained for independent grading.

5. Track-2 generalized closure ledger (April 2026, implementation milestone)
   Agent invocation, governance admission, closure event, and substrate consumption reduced to append-only records with explicit authority separation.

SYSTEMS AND TOOLS
Linux, systemd, Python, SQLite, Git, NATS, REST APIs, JSON, Ollama, DGX Spark, Cloudflare, Ed25519, x402, Base Sepolia, HTML, CSS, JavaScript.

ROLE FIT
- Forward-Deployed AI Engineer
- Agent Systems / Agent Infrastructure Engineer
- AI Evals and Reliability Engineer
- Applied AI Systems Architect
- AI Governance Engineer
- Technical Product Lead for Agentic Systems

OPERATING PRINCIPLES
- Worker self-report is not closure.
- Stored memory is not learning until a later attempt changes.
- Changed behavior is not improvement until a held-out result verifies it.
- Improvement does not grant authority.
- Historical documents are evidence inputs, not current runtime truth.
- Negative and undecidable results remain first-class records.
- Prefer the smallest intervention that closes the actual loop.

AVAILABILITY
Open to employment, contract engineering, design-partner work, and bounded agentic-AI closure and steering audits.


CURRENT COMPUTE ARCHITECTURE
- Qwen 27B local language target; exact package identity pending verification.
- Grok external high-compute worker; authority none.
- EVOKE visual-world candidate; isolated benchmark pending; authority none.
- SAGE sole governor of accepted state, admission, verification and closure.