🎬 Complete Portfolio: 25 Production Software & AI Systems

The showreel covers 25 production systems across four pillars: Autonomous B2B & AI SaaS (Partenero, Qbits, GEO-Fix, VeRobots.net, Tenders CRM, SereniCapelli.ro), Cybersecurity & Cloud Infrastructure (VeroShield, VeroPanel, VeroMail, VSM, VeroPress), Agentic AI Tooling (NEX IDE, Netor, The *er Group, Nexus Enterprise, Nexor, Telecomanda, Token Transformer, System-One Router), and Hardware, Robotics & Research (MUnlocker, SIMbox Explorer, VeroReason, Artificial Subconscious, Urban Traffic Simulator, Mobile Games & Media Server).

Yes. Augustin Olarian provides custom enterprise deployments, single-tenant/air-gapped installations, B2B architectural consulting, and fractional CTO leadership.

Partenero runs the Only Possible Customers AI gate on every discovered company website before initiating contact, ensuring messages are sent exclusively to businesses with a verified operational need for the offer, paired with multi-domain warmup and adaptive pacing.

It builds an executive dossier containing the prospect core services, likely operational bottlenecks, key decision-maker roles, and recommended communication tone.

Standard chatbots only generate text replies. Qbits connects six specialized agents through secure MCP bridges directly into authorized servers, workstations, browsers, and email inboxes to execute and verify real engineering deliverables.

Every agent action passes through a governance layer with scoped permissions, command allowlists, audit logging, and human-in-the-loop escalation for high-impact operations.

Generative Engine Optimization (GEO) is the engineering discipline of structuring websites, entity graphs, and verifiable facts so AI engines like ChatGPT, Perplexity, Gemini, and Claude cite your brand as the authoritative answer. GEO-Fix automates this via 600+ checks, JSON-LD generation, and live LLM citation probing.

Unlike static audit tools, GEO-Fix generates ready-to-deploy JSON-LD schemas, AI crawler rules, and factual answer capsules that are applied and verified directly in production.

Traditional WAFs look for known attack signatures and fail against novel zero-days. VeroShield verifies continuous behavioral and transport-level human authenticity at the gateway (VeroShield OS) and correlates it with local process/file guards (VeroShield Child), stopping automated scanners and exploit payloads before they can trigger the underlying vulnerability.

Yes. The enterprise edition is built for government and corporate infrastructures, while civil editions are released free on veroshield.org.

Legacy panels consume 2 GB to 4 GB of RAM, introduce large third-party attack surfaces, and charge escalating per-account licensing fees. VeroPanel runs in under 30 MB of RAM as a single compiled Go binary with zero licensing fees and strict per-tenant isolation.

Yes, VeroPanel provisions isolated per-domain PHP-FPM pools alongside integrated database management, DNSSEC, SSL, and backup orchestration.

Because VeroMail never uses or stores static passwords, there is no password for an employee to type into a phishing page or for an attacker to brute-force. Every session is verified via asymmetric cryptographic keys bound to authorized devices.

Yes. VeroMail exposes scoped, auditable API hooks for AI agents with strict rate and action guardrails, without ever handing IMAP/SMTP passwords to the agent.

Before applying any configuration or package change, VSM captures a reversible state snapshot and baseline HTTP probes for every hosted domain, then re-probes all endpoints after reload and automatically rolls back if any new 5xx regression is detected.

Yes, VSM manages bare-metal hypervisors, Linux web/database nodes, and Windows workstations/servers from a unified dashboard.

VeroPress does not execute interpreted third-party PHP plugins or allow arbitrary script uploads in web roots. All CMS, e-commerce, and SEO/GEO modules are compiled directly into a strongly typed Go binary with strict parameterized queries.

Yes. A single VeroPress control plane manages hundreds of distinct domains, themes, and localized content pipelines simultaneously.

Standard coding agents repeatedly stuff entire repositories and bloated chat histories into every prompt. NEX IDE maintains hierarchical index maps, per-file symbol headers, and lossless conversation compression so models read only the exact lines required.

Yes. You can switch providers mid-conversation or run multi-model brainstorming rounds where Anthropic, OpenAI, and Google models inspect the same workspace state and synthesize the strongest engineering solution.

Standard Chrome profiles spawn heavy duplicate processes and apply proxy settings only to an entire browser window. Netor isolates IP addresses and session stores per virtual browser, per tab group, or per individual tab inside a single window while using 10x to 50x less RAM and supporting native multi-agent AI control.

Yes. Netor continuously synchronizes tabs, notes, and active AI agent states across your PCs and mobile phone.

Through a single *er account, an AI agent can orchestrate tasks simultaneously across live Chrome/Netor browser sessions (Chromer), Windows workstations (Windowser), Linux servers (Linuxer), and Android smartphones (Androider).

Because these bridges grant deep operational control across computers, servers, browsers, and phones, access to The *er Group is provided strictly by special request and vetted B2B deployment.

You simply provide your domain URL. VeRobots crawls your pages, product feeds, and policies, cleans and chunks the content while preserving semantic structure, and indexes it into an isolated tenant RAG vault.

Every incoming message passes through behavioral rate limiting, prompt-injection sanitizers, and a cross-store e-commerce fraud risk engine.

Safe Chat acts as an on-premise cryptographic interceptor. Before a prompt leaves the institution network, all personal names, account numbers, contract values, and confidential entities are replaced with synthetic tokens (e.g., [ENTITY_A], [AMOUNT_1]). The external model reasons over the structure, and Safe Chat swaps the real values back in locally.

Yes. In air-gapped mode, Nexus Enterprise operates 100% on local GPU clusters with zero external network egress.

Lightweight models fail on complex tasks primarily because they lose track of multi-step planning and edge-case rules. Nexor externalizes the frontier model planning, security constraints, and verification scripts into a visual state machine that guides the lightweight model one verified step at a time.

Nexor excels at full-stack software architecture, public procurement tender compliance, and European Union grant engineering where hundreds of strict rules must be satisfied without a single omission.

Each user authenticates into an isolated OS and workspace context with dedicated private memory, while shared projects and company documentation reside in permission-governed common vaults accessible to authorized team members and managers.

Yes. While kept private rather than offered as a public SaaS, Augustin Olarian deploys and customizes the complete multi-user voice workspace architecture for enterprise clients.

No. It is not model routing or a third-party proxy. You run the exact frontier model (such as Claude Opus or GPT) directly from the provider, converted from standard subscription capacity into programmatic API keys.

To protect the proprietary architecture, internal details are not published publicly, but Augustin Olarian can deploy it privately on request or sell the complete transformer with its full source code.

When a giant frontier model executes 10 sequential tool calls over a 100k-token context, every turn incurs massive prefill latency and token cost. System-One runs the file-gathering and verification turns on ultra-fast models in milliseconds and invokes the frontier model once on a verified, compressed payload.

An independent verifier stage compares the compressed payload against the original source for numbers, negations, paths, and constraints before the frontier model is ever called.

Within 60 seconds of a tender being published, Tenders CRM parses the entire archive and highlights hidden eligibility blockers, margin-destroying SLA clauses, or vendor-locked specifications so executives only bid on profitable, winnable contracts.

Yes. It automatically drafts legally grounded clarification questions and builds a verified 1-to-1 compliance matrix mapping every requirement in the specification to the submitted dossier.

The AI agent governs customer consultation on the website, WhatsApp, and email, order verification, invoicing, warehouse coordination, courier AWB tracking, delivery exception handling, refill outreach, and promotional campaigns.

Yes. Augustin Olarian implements custom autonomous e-commerce and ERP/warehouse orchestration agents for established online brands.

Before writing any partition, the utility reads the hardware board revision from the bootloader, backs up device-unique radio calibration partitions, and verifies the cryptographic signature of the matching firmware image.

Technicians do not type IMEIs or fault logs manually; plugging the phone in over USB populates the cloud ticket, attaches the 60-point diagnostic report, tracks parts, and generates the customer warranty.

High-level wrapper libraries frequently lock up under multi-port concurrency and mishandle multipart Unicode encoding or delivery receipts. Driving the serial ports directly via asynchronous AT state machines and raw PDU frames guarantees maximum throughput and instant fault recovery.

SIMbox Explorer continuously monitors signal strength and USSD balances, rerouting queued messages to healthy ports in milliseconds and resetting stalled modems automatically.

An LLM never drives robot joints directly. VeroReason uses a strict three-agent consensus where the proposed trajectory must pass a deterministic inverse-kinematics physics solver and a hard-coded collision/torque safety auditor before motion is permitted.

Traditional industrial robot reprogramming takes 2 to 6 weeks whenever a part geometry changes. VeroSpatial and VeroReason perceive the new part in 3D and synthesize a verified trajectory in under 2 seconds.

Standard LLMs treat every prompt as an isolated conscious calculation, forcing them to re-read massive context windows and frequently forget operational rules mid-task. The Artificial Subconscious introduces pre-conscious associative filtering and instinctive safety reflexes before main inference begins.

Medulla separates transient conversational filler from durable decisions, numbers, negations, and causal lessons, verifying coverage independently before archiving raw logs.

Games use simplified agent heuristics and despawn stuck cars. The Olarian Urban Traffic Simulator models real municipal road graphs, exact legal traffic rules, and live ANPR camera feeds to solve real-world metropolitan congestion without despawning vehicles.

By recognizing plate vectors across consecutive intersections, the AI knows the exact volume heading toward each arterial exit and opens synchronized outbound green waves while recommending 15-minute departure offsets by district.

Most free children games on public app stores are packed with manipulative ad SDKs, behavioral trackers, and aggressive microtransactions. Our 34-game suite runs 100% offline at 60 FPS with zero ads and zero tracking.

It operates with zero monthly subscription fees, complete family privacy, live GPU hardware transcoding, automatic subtitle synchronization, semantic search, and unrestricted offline travel downloads.