Public Profile

KALIN PENKOV

AI SYSTEMS ARCHITECT | AGENTIC & LOCAL-FIRST AI | PRODUCT BUILDER

Bulgaria
Email: [add email]
GitHub: [add GitHub]
LinkedIn: [add LinkedIn]
Portfolio: [add portfolio link]


PROFESSIONAL PROFILE

AI Systems Architect and product builder with hands-on experience in agentic architectures, local language models, structured memory, autonomous workflows and controlled tool execution.

My involvement with artificial intelligence began years before the modern generative-AI wave, through early exposure to neuroevolution and evolutionary neural networks around Peter Chervenski, creator of MultiNEAT and MultiNEAT2.

My professional path developed through WordPress, PHP, web development, hosting infrastructure, databases and production deployment. It later expanded into local-first AI, agent orchestration, tool calling, persistent memory and safety-oriented execution.

Today, I design complete AI systems: defining components, responsibilities, communication contracts, autonomy boundaries, validation, failure recovery and future extensibility.


CORE AREAS

- AI systems architecture
- Agentic and multi-agent systems
- Local-first and private AI
- Local language models
- Structured and persistent memory
- Tool and function calling
- Workflow orchestration
- Schema-driven systems
- Model-provider abstraction
- Human-in-the-loop control
- Policy validation and permissions
- Logging, observability and recovery
- Product architecture


PROFESSIONAL EXPERIENCE

INDEPENDENT AI SYSTEMS ARCHITECT AND PRODUCT BUILDER

I design local and hybrid AI systems, agentic workflows and applications with practical business and creative value.

My work includes:

- agent, skill and workflow architecture;
- local models and provider abstraction;
- structured outputs and memory systems;
- tool execution and safety boundaries;
- logging, validation and failure recovery;
- product roadmaps and technical documentation;
- coordination of AI coding agents.


WEB STUDIO PENKOV — FOUNDER AND WEB DEVELOPER

Experience across the full lifecycle of web projects:

- WordPress and CMS architecture
- PHP, HTML, CSS and JavaScript
- MySQL and MariaDB
- cPanel, DNS, SSL and email
- migration and deployment
- caching and performance optimisation
- troubleshooting and maintenance
- client communication and technical planning

This experience gave me a strong production perspective: a system is not complete when it generates an output, but when it can operate reliably, remain maintainable and recover from failure.


SELECTED PROJECTS

SOVERIS

A local-first AI business ecosystem combining a desktop CRM, Android application and WordPress integration.

Core capabilities:

- universal business ontology;
- unified identities across multiple business contexts;
- contacts, tasks, agreements, offers, projects and payments;
- natural-language retrieval;
- local AI providers;
- voice input;
- offline queues;
- secure synchronisation;
- multi-business and multi-site support;
- retries, idempotency and logging.


SITEFORGE

An agentic system for producing complete web projects.

Architecture:

- Agents — specialised roles
- Skills — reusable procedures
- Workflows — production pipelines
- Schemas — output contracts
- Registry — discoverable system map

The system covers business analysis, design, image generation, website assembly, CMS architecture, security validation and iterative correction.

SiteForge is a case study in agentic-system design rather than merely a website generator.


CINEMA PERSONA STUDIO

A local AI filmmaking system for creating short films with persistent personas and visual continuity.

Components include:

- screenwriter and director agents;
- persona and scene management;
- shot fragmentation;
- model selection;
- timeline construction;
- start-frame and last-frame continuation;
- local video backends;
- continuity and final assembly.

The project demonstrates orchestration of generative components under resource, continuity and production constraints.


SOVERIS DEPLOY

An AI-assisted deployment platform designed to transform ZIP projects or repositories into production-ready websites.

Central principle:

Intelligent analysis with deterministic and controlled execution.

Architecture includes:

- project detection;
- structured deployment plans;
- policy validation;
- allow-listed actions;
- dry-run mode;
- backups;
- health checks;
- automatic rollback;
- human approval for destructive operations;
- audit logging;
- provider capability profiles.

Planned support includes cPanel, LiteSpeed, Apache, Nginx, PHP, MySQL/MariaDB, WordPress, static sites, React/Vite, SSL and cron.


SOVERIS SEED AND CORTEX

An experimental cognitive architecture focused on persistent memory, autonomous learning, uncertainty management and provenance.

It follows the Frugal Learning principle:

Observation
→ error or uncertainty
→ self-check
→ evidence
→ episodic memory
→ temporary adaptation
→ repeated validation
→ stable knowledge

Research areas include:

- episodic and structural memory;
- provenance tracking;
- confidence calibration;
- contradiction detection;
- competing hypotheses;
- targeted questioning;
- behavioural adaptation;
- controlled parameter updates;
- long-term continuity.


TECHNICAL SKILLS

AI AND ARCHITECTURE

AI systems architecture, agentic systems, multi-agent orchestration, local LLMs, structured memory, retrieval, tool calling, structured outputs, workflows, schemas, provider abstraction, safety policies, permissions, validation, logging and recovery.

SOFTWARE AND WEB

Python, PHP, HTML, CSS, JavaScript, WordPress, REST APIs, MySQL, MariaDB, cPanel, Apache, Nginx, LiteSpeed, SSL, DNS, hosting, deployment and migrations.

PRODUCT AND SYSTEM DESIGN

Product architecture, modular design, plugin systems, interface contracts, business ontology, offline-first architecture, mobile/desktop integration, risk analysis, technical documentation and roadmap planning.


TRAINING AND CERTIFICATIONS

- Python Image Recognition: Hands-On Artificial Intelligence Course — BitDegree, 22 March 2019
- Learn PHP Online — BitDegree, 18 February 2019
- Learn Web Design — BitDegree, 23 January 2019
- Cool CSS Tricks for Beginners — BitDegree, 20 January 2019
- How to Create dApps on Polygon — Learnoverse / BitDegree
- FIO: The Ultimate Blockchain Usability Layer — Learnoverse, 20 February 2023


WHAT DISTINGUISHES MY WORK

I combine AI foundations, production engineering and practical construction experience.

Web development and hosting infrastructure taught me reliability, maintainability and failure recovery.

Practical craftsmanship taught me sequencing, dependencies and quality control: every stage must be completed and validated before the next one begins.

I apply the same logic directly to agentic workflows and AI orchestration.


PROFESSIONAL OBJECTIVE

I am interested in roles and projects involving:

- AI Systems Architecture
- Agentic AI Development
- Local and Private AI
- AI Product Engineering
- Applied AI Research
- Workflow Automation
- AI-enabled Business Systems
- Architecture Consulting

I am particularly interested in opportunities where I can contribute not only to individual features, but to the architecture of the complete system.


GITHUB AND PORTFOLIO

My GitHub portfolio presents the projects as technical case studies through:

- architecture descriptions;
- diagrams;
- screenshots;
- demo videos and GIFs;
- agents, skills and workflows;
- technical decisions;
- safety mechanisms;
- limitations, roadmaps and development status.

Source code may remain private because the systems contain proprietary work.