# Jernej Teraš

AI Native. Product Developer. Human First.

Personal CV and portfolio
The human behind the work: projects, path, and method.
Nothing is random.

## About
I can build complete products on my own — vision, design, architecture, and execution all held by one person, with no handoff and no translation loss between the idea and the build.

My edge is range. A decade of self-employment across human performance, financial markets, and technology gave me a way of thinking most specialists never develop: I work in systems and in detail at the same time, holding the big picture without losing the small things. Before AI, that breadth was hard to put to use in one person. Now it's my whole advantage — one mind moving across product strategy, design, engineering, and market judgment without losing the thread.

I work fast and I hold a high standard. I learn quickly, adapt quickly, and stay current with where the technology is going. Quality isn't negotiable for me — nothing ships until it's right. The technical execution can be offloaded; the vision, the judgment, and the taste can't.

I trust my product instinct, and I'd rather let the work prove it than oversell it. The proof of work is in what I've built.

## Portfolio Projects
## LocalAI
Slug: local-ai
Status: In Development
Visible in portfolio: yes
Portfolio order: 3
Tagline: Your own private AI — runs entirely on your machine, no cloud, no data leaving your desk.
Idea:
Every major AI chat tool sends your conversations to someone else's server. For anyone working with sensitive information — or anyone who simply values privacy — that's a dealbreaker. LocalAI was built to solve that: a full-featured AI chat app that runs powerful language models directly on your own hardware, completely offline, with nothing phoned home. Cloud mode is available too, via your own API keys, keeping you in control either way.
What it does:
LocalAI runs as a desktop app on Windows with an NVIDIA GPU. You point it at your collection of GGUF model files and it handles everything else — loading the model, managing memory, serving responses. The chat interface feels like any modern AI assistant: streaming responses, markdown rendering, conversation history, and a clean light/dark UI.

Local and cloud modes are fully isolated. In local mode, models run on your GPU via llama.cpp — fast, private, free after setup. In cloud mode, you connect any OpenAI-compatible provider (DeepSeek, Groq, OpenRouter, or any custom endpoint) using your own API key. Each mode keeps its own separate conversation list so nothing mixes.

Every conversation is saved automatically. Per-chat system prompts let you set instructions that persist across sessions. A context meter shows token usage against the model's limit in real time, colour-coded so you know when you're approaching the edge. If you reopen an old conversation that was created with a different model, the app warns you and lets you decide how to handle it.

API keys are encrypted at rest — never stored in plain text. All data lives locally as JSON files on your own machine.
What's left:
Core chat functionality is complete. Remaining work: a one-click launcher so non-technical users don't need to start two terminal windows, friendly model display names instead of raw .gguf filenames, per-message retry and edit, voice input, a prompt library, tool calling, and a proper Windows installer. The app currently requires manual setup — packaging it for easy distribution is the main thing standing between this and a wider release.
Business model:
One-time purchase targeting privacy-conscious professionals, developers, and power users who want local AI without the setup complexity. A hosted-key tier is also possible — users pay for access without needing their own API key or GPU.
Tech stack:
- React
- Vite
- Python (FastAPI)
- llama.cpp (CUDA)
- GGUF model format
- OpenAI-compatible API (BYOK)
- Fernet encryption
- Windows / NVIDIA GPU
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## Lynqd
Slug: lynqd
Status: In Development
Visible in portfolio: yes
Portfolio order: 4
Tagline: Local-first AI outreach engine — personalized emails and proposal PDFs, no cloud, no accounts.
Idea:
Cold outreach is time-consuming not because the work is hard, but because personalizing every email and proposal from scratch is. Lynqd was built to solve that for solo operators and small teams: set up your business context once, configure your tone and services, and generate a tailored email and branded PDF proposal for any prospect in seconds. Everything runs in the browser — no backend, no accounts, no data leaving your machine.
What it does:
Lynqd is built around Presets — reusable sender profiles that store your identity, tone, strengths, pricing, CTA, and prompt templates. For each outreach, you pick a preset, enter or select a client, choose your output mode (email only, or email plus PDF proposal), and hit generate. The app combines your preset context with the client details and calls your configured AI to produce a personalized result.

The result is fully editable. You can regenerate the entire output, regenerate just the email, or select any section of the PDF and ask the AI to redo just that part with optional instructions. Once you're happy, copy the email, open it directly in your mail client, preview the PDF, or download it.

Business configuration covers everything: branding (logo, colors, background image for PDFs), a services and products library, and a Knowledge Base where you can inject company context as text or upload files (.txt, .pdf, .docx). A Business Overview mode pitches your entire catalog in a single generation without selecting individual services.

The social engine is a separate module for generating X (Twitter) posts and threads — create social profiles with tone and context, pick a post type, and generate ready-to-copy content.

All data lives in IndexedDB in the browser. A full backup and restore system lets you export everything to a single .lynqd file and import it on any machine. No accounts, no sync, no cloud dependency.
What's left:
V1 is complete and working. V2 is actively in progress — dark design system rollout across remaining pages, social content engine polish, and generation quality tuning. Planned next: Tauri desktop packaging so non-developers can run it without a terminal, a hosted-key tier for users without their own API access, direct email sending via SMTP, and hosted shareable PDF links with open tracking.
Business model:
One-time purchase for the local app. A hosted-key tier adds recurring revenue — users pay a monthly fee and Teras AI provides the API access. Direct email sending and hosted PDF links are natural upsell features for a higher tier.
Tech stack:
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- shadcn/ui (Radix UI)
- Dexie (IndexedDB)
- TanStack Query
- OpenAI-compatible LLM (BYOK)
- @react-pdf/renderer
- Zod
- Tauri (planned)
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## DevHub
Slug: devhub
Status: In Development
Visible in portfolio: yes
Portfolio order: 6
Tagline: Encrypted local vault for developers — credentials, subscriptions, billing, and projects. No cloud, no accounts.
Idea:
Every developer accumulates dozens of service credentials, API keys, and subscriptions spread across browser bookmarks, notes apps, and memory. DevHub was built to fix that — a single encrypted vault that lives entirely in the browser, secured by a master password, with no backend and nothing leaving your machine. The billing tracker came from a personal need: actually knowing what you're spending every month across all your dev tools and services.
What it does:
DevHub is a local-first encrypted vault that unlocks with a master password. Everything inside is AES-GCM encrypted using a key derived from your password — credentials, notes, billing data, all of it. Nothing is ever sent anywhere.

The Services library stores credentials for every tool and service you use — URL, login details, category, billing type, and notes. Paid services get a full billing history: initial payment, renewal tracking, overdue alerts, and upcoming payment reminders for anything due within 7 days. The dashboard shows current month spend, year-to-date spend, active service counts, and a full payment timeline. A tax and bookkeeping CSV export covers everything.

The Emails library manages email accounts with passwords, phone numbers, and notes — all encrypted. Projects link to your existing services so you can see exactly which tools a project depends on, with a tech-stack snapshot workflow that generates a structured prompt you can paste into any AI for documentation.

The entire vault exports to a single encrypted .devhub backup file. Import it on any machine, enter your password, and everything is restored. The app also auto-exports a backup silently after every write so you never lose data.
What's left:
Core v1 is complete and in daily personal use. Remaining work: TypeScript migration (deferred to v2), desktop packaging via Electron or Tauri, and a store listing. Multi-device sync is intentionally out of scope — the local-first model is a feature, not a limitation.
Business model:
One-time purchase via the Teras AI store. Target: developers and solo operators who want a private, offline-first alternative to password managers and subscription trackers. No hosting costs, no backend — near-zero marginal cost per sale.
Tech stack:
- React
- Vite
- Zustand
- Tailwind CSS
- Web Crypto API (AES-GCM, PBKDF2)
- localStorage
- Tauri / Electron (planned)
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## TAO Intel
Slug: tao-intel
Status: Paused
Visible in portfolio: yes
Portfolio order: 7
Tagline: Bittensor intelligence platform — live market data, trading analytics, and AI-powered insights for TAO holders and validators.
Idea:
The Bittensor/TAO ecosystem had no single place to track everything that mattered. Holders, validators, and subnet operators were juggling multiple tools just to get a basic picture of the market. TAO Intel was built to replace all of them — one dashboard with live market data, trading analytics, a curated knowledge base, and context-aware AI analysts that answer questions using real platform data rather than generic knowledge.
What it does:
TAO Intel is organized into three live dashboards. The Market tab shows current TAO price, performance across multiple timeframes, and a candlestick chart with volume pulled from CoinGecko. The Trading tab surfaces TAO/USDT and TAO/BTC order book depth, ticker stats, buy/sell volume analysis, and volume profile with POC/VAH/VAL levels from Binance. The Learn tab is an admin-curated knowledge base of guides, FAQs, and glossary entries.

Every tab has its own context-aware AI analyst. The AI system is fully profile-driven — prompts, token budgets, context behavior, and tab availability are all admin-configurable rather than hardcoded. Admins can create multiple AI personas per tab, test them live, set defaults, and switch between providers (Gemini, Groq, Chutes/DeepSeek).

Behind the scenes, a background data pipeline handles all ingestion: scheduled CoinGecko and Binance jobs with retry logic, exponential backoff, job execution tracking, and failure alerting. Market snapshots and daily closes are stored separately so real-time stats and historical charts never pollute each other.
What's left:
Market, trading dashboards, AI chat, AI profile management, and knowledge base admin are all working. Development is currently paused. Remaining work includes user authentication and JWT, public Learn content rendering, Bittensor on-chain data integration, and paid subscription tiers. The foundation is production-grade and ready to resume if there is sufficient interest in the TAO ecosystem.
Business model:
Freemium SaaS. Free tier with limited AI queries per day. Premium (~€3/month) for unlimited queries. Pro (~€10/month) with API access. Target audience: TAO holders, validators, and subnet operators who need consolidated analytics without juggling multiple tools.
Tech stack:
- React 18
- TypeScript
- Vite
- TailwindCSS
- TanStack Query
- lightweight-charts
- Node.js
- Express
- PostgreSQL
- node-cron
- CoinGecko API
- Binance API
- Gemini 2.5 Flash
- DeepSeek (Chutes)
- Groq
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## CoachOS
Slug: coach-os
Status: Paused
Visible in portfolio: yes
Portfolio order: 8
Tagline: Fitness trainer amplification platform — one trainer, many clients, zero dropped balls.
Idea:
A good personal trainer has a system. The problem is that system lives in their head, in spreadsheets, and in WhatsApp threads — which means it breaks down past a handful of clients. CoachOS was built to give trainers a proper operating system: a single platform where they build workout plans, track client progress, communicate, and let an AI assistant handle the questions they answer fifty times a week.
What it does:
CoachOS is a full trainer-client platform with two separate sides. The trainer side is a complete admin system: build weekly workout plans per client with exercises, sets, reps, rest times, and supersets; manage an exercise library with categories, difficulty levels, form tips, and media; assign video workouts from a curated YouTube library; create questionnaires for client onboarding and check-ins; and track measurements, daily metrics, and progress photos per client.

The client side is a clean dashboard built around today's workout. A focus view shows exactly what needs to be done — no noise. A week view gives the full picture. Clients mark workouts complete, leave feedback with difficulty and enjoyment ratings, and see their history and progress charts over time.

The AI assistant is trained on the trainer's specific methodology, not generic fitness knowledge. It has full access to the client's profile, assigned workouts, measurements, and questionnaire responses — so it answers like an extension of the trainer, not a generic chatbot.

Trainer-client messaging is built in with real-time updates. The platform supports English, Czech, and Slovak with a translation system that handles both static UI strings and dynamic content through DeepL.
What's left:
Core workout planning, client dashboards, exercise library, video library, measurements, metrics, questionnaires, AI assistant, messaging, and multilingual support are all implemented. Remaining work: Stripe payments and subscriptions, AI-guided client onboarding flow, production deployment, and a final QA pass.
Business model:
SaaS subscription for trainers. A trainer pays a monthly fee and can manage an unlimited number of clients on the platform. Pricing scales with the number of active clients or as a flat monthly rate. Low marginal cost per additional client makes this high-margin at scale.
Tech stack:
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- Prisma
- PostgreSQL
- NextAuth (Auth.js v5)
- Cloudinary
- TanStack Query
- DeepL API
- Chutes AI (DeepSeek-R1)
- Upstash Redis
- YouTube Data API
- TipTap
- Recharts
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## Degen Academy
Slug: degen-academy
Status: Paused
Visible in portfolio: yes
Portfolio order: 9
Tagline: Gamified crypto education platform — learn blockchain, earn ranks, and compete on a global leaderboard.
Idea:
Crypto is genuinely hard to learn because most resources are either too shallow or assume too much. Degen Academy was built to fix that — a structured learning platform where you work through topics conversationally with an AI tutor, prove your knowledge through practice exams, and earn a rank that reflects what you actually know. The gamification is not cosmetic — mastery and rank update from real learning outcomes.
What it does:
Degen Academy is a structured crypto education platform built around AI-guided learning sessions. You pick a topic — Blockchain Basics, Wallets, Security, Tokens, Charts and Trading, DeFi, NFTs, Smart Contracts, Centralized Exchanges — and dive in. Each session is a conversation with an AI tutor that adapts depth based on your responses, asks follow-up questions, and generates a learning summary when the session completes.

Keyword mastery tracks your progress across topics through a shared keywords model — as you demonstrate understanding, mastery scores update automatically across ten progression tiers. A global leaderboard ranks all learners by mastery outcomes so you can see where you stand.

Practice exams test retention with persisted attempts and aggregate scoring maintained in the database. The AI tutor runs on Gemini 2.5 Flash with per-user rate limiting enforced server-side through edge functions.

Authentication uses a PIN-based system with edge-function hashing. The admin side is a full CMS for managing categories, keywords, questions, and knowledge points — with bulk .txt import and export, diff-style preview, and drag-and-drop reordering.
What's left:
The core learning platform is complete. Development is paused. The project was built on Lovable with Lovable Cloud infrastructure — it is intact and resumable but not currently being actively developed.
Business model:
Freemium or one-time access model targeting crypto newcomers who want structured education rather than scattered YouTube videos. Potential for premium topic packs or certification tiers.
Tech stack:
- React 18
- TypeScript
- Vite
- Tailwind CSS
- shadcn/ui (Radix UI)
- React Router
- TanStack Query
- Lovable Cloud (Auth, PostgreSQL, Edge Functions)
- Gemini 2.5 Flash (Lovable AI Gateway)
- Recharts
- bcryptjs
- @dnd-kit
- Lovable Hosting
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## Teras AI
Slug: teras-ai
Status: Live
Visible in portfolio: yes
Portfolio order: 10
Tagline: The business behind the work — company and portfolio site for a build, consult, and education operation.
Idea:
Every project needed a home base — a place that presents the services, showcases the products, and makes it easy for potential clients to understand what Teras AI does and how to work together. terasai.dev is that site: a premium dark portfolio and business site built to the same standard as the products it represents.
What it does:
Teras AI presents four service areas — Build, AI Integration, AI Consulting, and AI Education — each with dedicated pages and sub-pages covering specific offerings. Build covers landing pages, web applications, and desktop and mobile apps. AI Integration covers local and private AI, chatbots and assistants, workflow automation, and custom AI features. AI Consulting offers an AI audit, strategy sessions, and ongoing advisory. AI Education covers seminars, workshops, and one-on-one sessions.

The portfolio section showcases all in-house products with a cinematic carousel, per-project detail pages with image galleries, and admin-controlled visibility so projects can be staged for release without removing them from the data layer.

The store hosts product listings with full detail modals and image carousels, wired to Gumroad for purchases when products go live. A contact modal handles inbound enquiries via Netlify Forms.

The site uses a custom full-page scroll system with section transitions, a mobile-specific navigation with boundary-based section scrolling, animated hero particles with pointer attraction, rotating taglines, service card media, and an About section carousel. All content lives in typed data modules — no hardcoded text in components.
What's left:
The site is live and complete. Remaining work: activating the Dev Hub store listing with real purchase URLs, a blog section, CMS integration for easier content updates, analytics, and Slovenian translations. The contact form depends on Netlify Forms configuration for production email delivery.
Business model:
This is the company site — it generates leads and client enquiries for the Build, Consulting, and Education services. The store adds a direct product revenue channel. No recurring costs beyond hosting.
Tech stack:
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- Framer Motion
- Netlify
- Netlify Forms
- Gumroad
- next-themes
Links:
- Live Site: https://terasai.dev
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg
## Jernej Teraš — CV Site
Slug: jernej-teras-cv
Status: Live
Visible in portfolio: yes
Portfolio order: 11
Tagline: Personal CV and portfolio site — the human behind Teras AI, built to be read by people and fetched by AI.
Idea:
The Teras AI business site shows what the company does. This site shows who built it. A personal CV site that goes beyond a PDF — full project portfolio, personal story, capabilities, and methodology, all in a cinematic dark format that reflects the same standard as the products it documents. Built to be fetchable by AI so anyone can point a tool at the URL and get a complete picture.
What it does:
The site is organized into four sections plus a landing screen. About presents current positioning and background. Capabilities covers five areas — Methodology, Project Types, Technologies, Crypto and Finance, Human Performance — each expandable into a full detail view. Projects is a full-screen carousel of every product built, each with a detail page covering the idea, what it does, tech stack, and current status. Story is a five-slide cinematic presentation covering the personal journey: Origins, Sport, Training, Crypto, and AI and Building.

Navigation is non-linear — visitors move freely between sections rather than scrolling through a fixed sequence. Desktop uses a custom full-page scroll system with fade-to-black transitions. Mobile uses a separate component tree with boundary-based section transitions and swipe navigation.

All content lives in a single typed data file. A password-protected admin panel controls project visibility and ordering. The site is bilingual — English built, Slovenian to follow. Built with SSR and semantic HTML so the full content is readable by AI scrapers and fetch tools without JavaScript execution.
What's left:
Phase 1 shell and all personal content are complete. Phase 2 is ongoing — adding real project entries one by one, real project screenshots, Slovenian translations, wiring the contact form, and linking from terasai.dev.
Business model:
This is a personal CV — it generates opportunities, not revenue. Its job is to represent the full body of work clearly and credibly to anyone who lands on it.
Tech stack:
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- Framer Motion
- Netlify
- next-themes
Links:
- Live Site: https://jernejteras.online
Gallery images:
- /images/placeholder-1.jpg
- /images/placeholder-2.jpg
- /images/placeholder-3.jpg

## Capabilities
### Methodology
Vision-led, AI-executed, human-judged.

I start with a clear product vision — what it needs to do, who it's for, and what makes it worth building. Before I write a single line of code, I investigate the problem, talk it through, and make sure I understand it. I make the architecture decisions upfront, consider the edge cases, and map the path before execution begins. This is how I avoid the expensive mistakes that come from building first and thinking later.
When a knowledge gap appears, I close it before moving forward — through research and using AI as a learning partner to build genuine understanding of unfamiliar territory. I don't skip past what I don't know. I stop and learn it.
From there, AI tools handle the execution layer — each one chosen deliberately for what it does best, connected and directed by my judgment. The skill is knowing which tool to use for what, how to prompt precisely, how to catch when the output is wrong, and how to steer back on course.
This isn't about typing less code. It's about thinking at a higher level — staying in the vision while AI handles the implementation. Underneath it is decades of self-directed computer literacy: I got my first computer at ten and never stopped — taking them apart, figuring out how servers work, how databases are structured, how systems connect. That foundation is what separates informed direction from blind prompting. The result is a solo building capability that would traditionally require a team.
### Project Types
categories of products built

Web App, Mobile App, Blockchain Contract, Video Platform, AI Integration, E-Commerce, Admin Dashboard.
### Technologies
frameworks, tools, platforms

Next.js, React Native, TypeScript, Supabase, Tailwind CSS, Framer Motion, Prisma, CosmWasm, Cloudflare R2, SQLite.
### Crypto & Finance
Advanced fluency across the full crypto ecosystem — self-taught, battle-tested.

Crypto is knowledge I carry completely. Not one corner of it — the whole thing. I'm fully fluent across the entire market and I've used all of it firsthand, with real capital and real consequences. This isn't knowledge I read about. It's knowledge I lived.
I know it from the ground up. Wallets, private keys, seed phrases, self-custody, and security. Setting up and moving across different chains, different wallets, different systems. Sending, receiving, swapping, bridging between networks. Centralized and decentralized exchanges. Staking, yield farming, liquidity provision, leverage, NFTs. The full stack of how this ecosystem actually works in practice — the parts most people never get comfortable with, I use without thinking about it.
And I understand what sits underneath. On-chain mechanics — reading what's actually happening on a network rather than guessing. Tokenomics — how a token is structured, where the supply sits, how the incentives are designed, and whether the thing holds up or quietly falls apart. Market structure — Bitcoin dominance cycles, total market cap flows, altcoin season patterns, how the whole board moves together. That's the difference between following hype and understanding the system.
I've been through more than one full cycle. I've made significant gains and taken real losses, and both sides taught me things no course could. The result is someone who can meet anyone wherever they are — whether that's their first wallet or the deep end of the market — and someone who sees the whole picture clearly.
That fluency already feeds into what I build. Crypto analytics platforms, market intelligence tools, token trackers, on-chain data products — built on real understanding of the system underneath, not a surface reading of it.
### Human Performance
A decade of applied human performance — body, mind, and recovery.

I've spent my whole life around the human body, and more than a decade working with it professionally — as an athlete, a coach, and a personal trainer. But the thing I keep coming back to is simpler and bigger than any of that: the human body is the most technologically advanced machine that exists. It's just technology that happens to be alive — it self-regulates, self-repairs, and runs a near-infinite number of systems and processes at once, all the time, without being told to.
Once you really get inside how it works, the understanding goes far beyond exercise programming. It's how the nervous system adapts before the muscle does. How nutrition, sleep, and stress interact with performance. How chronic pain patterns form and how movement corrects them. How recovery isn't the gap between the work — it's the foundation the work is built on. I've worked across strength training, athletic performance, sport-specific conditioning, and movement, in constant exchange with physiotherapy through a close professional relationship. The approach was always holistic: treat the person and the whole system, not just the program.
What I didn't expect is how directly that thinking carries over. One system is biological, the other is technical — but the deeper I go into building with technology, the more I see the same patterns. Inputs and outputs. Feedback loops. Bottlenecks. Things that look like one problem on the surface and turn out to be another underneath. Learning to read a living system taught me how to read a built one. The two are far closer than most people realize, and that's exactly why this background isn't a previous chapter — it's part of how I think about everything I build now.

## Story
A placeholder career arc: from years of personal training and basketball refereeing, through gym management, into a focused shift toward software, AI-assisted building, and product development.

### Origins
Born in Maribor, Slovenia, in 1990. From the very beginning, I was the kind of kid who had to know how things worked — taking things apart, figuring them out, asking why until there was no further why to ask. The curiosity never switched off. It still hasn't.
Two things were there from the start. One was sport — I did all of it, constantly, from as early as I can remember. Ping pong, soccer, volleyball, track and field, basketball. If it involved moving and competing, I was in. The other was technology. I got my first computer at ten and immediately started pushing at its edges — pulling it apart, breaking it, fixing it, learning what it could really do. Never formal, never for an audience. Just relentless curiosity that needed somewhere to go.
School came easily. I attended II. gimnazija Maribor as a Zois Scholar — Slovenia's state merit award for exceptional cognitive and academic achievement — and later started studying civil engineering. The formal path was never really the point, though. It was just the first place a restless, system-hungry mind had to put itself while the real direction took shape.
Looking back, none of it was random. Every thread that mattered later was already here at the start — the body, the machine, and the need to understand both completely.
### Sport
Sport was the first language I was fluent in. As a kid I played everything, but at fourteen basketball became the serious one. I joined a team and started training properly, and from there it was years of training, competing, and climbing — eventually reaching the second division of Slovenian basketball. Not professional fame, but a real level, earned the hard way, against people who were good.
Basketball taught me things I still use every day. How to read a situation fast and act before it fully resolves. How to perform when it actually matters and the pressure is real. What discipline feels like when no one is making you do it. How to lose, learn, and come back. A team sport at that level is a live system — constantly moving, everyone reacting to everyone, and you either see the whole floor or you get left behind.
Alongside playing came coaching — working with younger kids coming up through the game. That was the first time I realized I didn't just want to do the thing, I wanted to understand it deeply enough to teach it, to break it down and rebuild it in someone else.
By my early twenties the playing chapter was winding down. The question was never whether to stay in sport — it was how. And the answer was already forming: if the game itself fascinated me, the machine playing it fascinated me more.
### Training
The move into personal training was the most natural thing in the world. After years obsessing over physical performance, training methodology, and how the human body actually works, becoming a trainer wasn't a career change — it was just making it official. I got licensed and was working within months.
It moved fast. Inside the first year I was promoted to running and managing a center for one of Slovenia's gym chains — responsible for the place, not just my own clients. I did that for another year before making a bigger move: I left for Brussels to go fully independent. Six years self-employed in a foreign city, building a client base from nothing, operating entirely on my own. I came back to Slovenia in 2021, and I'm still training to this day — over a decade in, the chapter never really closed.
A decade of working for yourself teaches things no job can. How to operate without a safety net. How to sell yourself and deliver every single time. How to keep going when there's no one above you setting the direction — when the standard, the discipline, and the next move all have to come from you. That's not a side lesson. That's the exact mindset I build with now.
And underneath all of it, the real obsession was never just fitness. It was the body itself — the most advanced system I'd ever get my hands on. Learning to read it, adjust it, and improve it was the training ground for everything that came next. I just didn't know yet what the next system would be.
### Crypto
COVID changed the pace of everything. Suddenly there was time to go deep on something, and the thing that pulled me in was crypto. Once I was in, there was no shallow end — I went all the way down, and I stayed down there for years.
It became a second full-time obsession running right alongside the training. Countless hours of research, testing, and living inside the market day and night. I wasn't watching from the sidelines — I was in it, with real money on the line, learning the way you only can when the consequences are yours. Some of it paid off in ways most people never get to experience: genuine outcomes that only come from being early, doing the work, and holding conviction when everyone around you is doubting. And some of it went the other way — crypto is unforgiving, and the expensive lessons came with the territory. Both sides are real. I wouldn't trade either.
What the cycles really taught me wasn't about money. It was about systems, incentives, psychology, and timing — how a market made of millions of people actually behaves, and how to keep a clear head when fear and greed are running everything around you. I learned to sit with uncertainty, to think independently, and to be honest with myself about what I did and didn't understand. I'm still in it today — less as a trader chasing the next move, more as someone who simply understands this world and stays close to it.
The knowledge never left, and that was always the part that mattered. I genuinely believe we're entering a new chapter in what money is and how it works — and between the years inside the market and the fluency I built, I'm positioned to be right at the front of it. The same instinct that ran through everything before — read the system, understand it deeply, then do something with it — was about to find its sharpest expression yet.
### AI & Building
The obsession with AI didn't start when the building did. From the moment large language models became genuinely capable, I was in — prompting, testing the limits, figuring out what these systems could actually do versus what people claimed. The curiosity compounded fast, and within months AI went from a tool I was using to the foundation I wanted to build on.
I believe AI is the technology of the future — the single biggest shift of our lifetime. A lot of people don't understand it yet, and some of the fear around it is fair. But where others see a threat, I see possibility. I genuinely believe this technology can improve the human experience and make life better in ways we're only beginning to glimpse, and that belief is a big part of why I build. Not to chase a trend — to actually put it to good use.
And this is where everything finally comes together. For years it looked like I was jumping from one thing to another — sport, the human body, training, finance, crypto, self-directed learning. I now understand it was all preparation. The understanding of the human system, the fluency in complex markets, the relentless need to learn and connect things others keep separate — none of it ever fit a single lane, and that sometimes felt like a weakness. AI turned it into the whole point. The kind of mind that thinks in broad systems, stays curious, and refuses to narrow down is exactly the kind of mind this technology rewards.
The technical execution can be offloaded now. The vision, the judgment, the taste — that's the part that's mine, and it took a lifetime to build. In a short time I've built across mobile, web, blockchain, and AI integration. All solo. All from scratch. This isn't a career pivot. It's the moment everything I am was always heading toward — and I love it more than anything I've done before.

## Contact
hello@example.com