I build content systems, interactive web experiences, Shopify interfaces, AI-assisted tools, and production workflows where design and engineering overlap. My work ranges from ADG/AleDellaGiusta visual packaging connected to 149M+ measured views, to WebGL automotive prototypes, automation pipelines, e-commerce builds, key art production, and concept systems for Motor Valley.
I study Computer Engineering at Unimore in Modena and operate Studio Aiman as a practical lab: every project is either shipped, prototyped, or built to prove a specific capability. This portfolio keeps the previous minimal UX, but the case studies now explain the actual work behind the files — without exposing private client material.
Content systems · visual packaging · pitch material · production memory
This section is not an exposed file tree. The folders are the evidence layer; the public story is the workstream they reveal. Across the ADG/Ale archive, my role expanded from individual episode visuals into a repeatable production system: thumbnail direction, raw source research, PSD/PSB compositing, multilingual adaptation, creator identity assets, pitch material, merchandise support, analytics/proof decks, and Shopify-related work.
The important point is not that the archive contains many files. It is that the file structure shows a real production loop: messy source material became visual concepts, concepts became working compositions, compositions became export variants, and the strongest patterns became reusable campaign memory.
02DirectionChoosing the visual metaphor and character focus.
03BuildPSD compositing, typography, and contrast.
04AdaptLanguage variants and platform formats.
05ReuseStrong patterns kept as production memory.
This is the part a recruiter needs to understand: I was not simply making isolated graphics. I was helping maintain a system where story material, creator identity, commerce, pitch decks, and performance proof had to stay visually coherent over years.
02 What I Actually Did
Editorial packagingEpisode visuals from raw material to final thumbnail.Research, composition, versioning, review, and publishing support.
Creator brand layerMedia kit, banners, profile assets, tour material.Not a one-off logo exercise: it supported presentation across contexts.
Business layerPitch material, Kimi/Mercedes assets, proof decks.The archive shows creative work moving into business-facing communication.
Private deck pages, confidential strategy, and sensitive client material are intentionally abstracted. The portfolio explains the process and role without exposing what should remain private.
03 Production Memory
The phrase "production memory" matters because long-term creator work is not only about taste. It is about preserving what worked: title treatments, emotional hooks, source-image logic, language adaptation, pitch framing, launch assets, and proof that creative decisions were connected to audience response.
Decision
Do not treat every episode as a blank page. Build visual patterns that can be adapted.
Decision
Keep design and production close. The same system supports thumbnails, decks, merch.
Decision
Turn repeated friction into tools. Automate bottlenecks.
The gap
Pagani Configurator
16 January, 2026
Self-initiated automotive WebGL prototype
A high-fidelity WebGL design concept developed to configure bespoke specifications for models like the Utopia, Imola Roadster, and Huayra Codalunga in real-time browser pipelines. Built unsolicited as a portfolio piece targeting the automotive sector — no client, no brief, no deadline. Just a question: what would a Pagani configurator look like if it were built by someone who actually understands both Three.js and the brand?
01 Overview
Pagani Automobili builds fewer than 50 cars per year. Every unit is bespoke — material choices, color combinations, and carbon weave specifications are negotiated directly between the client and the atelier in San Cesario sul Panaro, twelve minutes from my desk in Modena.
The existing configurator experience for cars at this price point is usually a PDF sent by email, or at best a Flash-era web tool that hasn't been touched since 2019. The gap between the product and the digital experience is enormous.
The three models featured — Utopia, Imola Roadster, Huayra Codalunga — were chosen for their geometric contrast. Together they stress-test the rendering system across completely different silhouettes.
02 WebGL Render
The renderer runs entirely in the browser using Three.js and a custom WebGL pipeline. No server-side rendering, no pre-baked images. Every color change, every material swap, every camera angle is computed in real time at 60fps on a standard laptop GPU.
Camera positions are locked to three presets — FRONT, 3/4, SIDE — because free orbit on a car configurator is a UX mistake. Users don't know where to look. Locked angles reproduce how automotive photography actually presents vehicles.
03 Shader Architecture
To render the specific weave of lightweight composite fibers accurately without using large heavy texture maps, we designed a procedurally mapped clear coat shader. The carbon fiber weave is generated mathematically at render time using a grid of sinusoidal functions that reproduce the 2x2 twill pattern.
Dark UI on dark car. A white background configurator reads as a consumer tool. Dark reads as atelier.
Copy
Italian copy. The export button says "ESPORTA CONFIGURAZIONE PDF." Pagani is Modenese, the language signals context.
Pricing
No price. At this level of the market, price is discussed in person. Showing a number in a browser would be wrong.
Earth ADG
21 January, 2026
3D data metaphor · globe interface · ADG reach
A realtime 3D WebGL experience that maps worldwide user interaction and live content distribution routes onto an instanced spherical render. Built to visualize the geographic reach of a content pipeline generating hundreds of millions of views — and to answer a question no standard analytics dashboard could: where in the world is this content actually landing?
01 Concept
ADG has 149 million views. That number lives in a YouTube Studio dashboard as a flat integer. It tells you nothing about where those views are, how they cluster, which markets are growing, or what the geographic shape of the audience actually looks like.
The brief was self-assigned: make the number mean something spatially. Build something you can look at and immediately understand that this content reaches people across six continents. The metaphor of a globe was obvious. The execution was not.
02 Interactive
The sphere is fully interactive. Users can click any data point to surface the content associated with that geographic cluster. The tooltip system uses a raycaster to convert mouse coordinates to spherical surface intersections.
03 Technical Architecture
Converting latitude/longitude to 3D Cartesian coordinates on a unit sphere is straightforward trigonometry. The implementation challenge is doing it for thousands of data points at 60fps without dropping frames.
The main frame budget concern was texture memory. A 4K Earth texture at full resolution is 64MB uncompressed in GPU memory. On mobile, this is unacceptable. The implementation uses KTX2 compressed textures with Basis Universal transcoding. GPU memory footprint drops to under 8MB with no visible quality loss on screens below 1440p.
Aiman Studio is a production system built from scratch without visual site builders. The project serves as a live, functional sandbox to demonstrate custom performance layers and tight layout logic. The primary objective was simple: build a modular engine capable of rendering technical case studies synchronously while maintaining stable 60fps on mobile displays.
01 Overview
Most developer portfolios are one of two things: a Notion page with a list of GitHub links, or a Webflow template with a profile photo swapped in. Neither demonstrates anything about how the person actually builds.
The brief I gave myself was different. The portfolio has to be the proof. If I am claiming to build high-performance, design-precise web experiences for clients — then the portfolio itself must be a high-performance, design-precise web experience. Not a description of one. Not a screenshot of one. The actual thing.
02 Architecture
The frontend operates on top of a highly optimized React layer using Vite. Rather than introducing heavy external UI dependencies which cause layout shift, every structural element is custom-coded.
Routing is handled with React Router in lazy mode. Every route is a separate chunk. Chunks are pre-compiled and fetched asynchronously. To prevent micro-delays when clicking items, the application registers mouse hover intents to trigger instant pre-resolving — the chunk is in memory before the user clicks.
router.ts
// Preload on hover intent, not on click
const preload = (page: string) => () => import(`./pages/${page}`);
<Link
to="/case-studies"
onMouseEnter={preload('CaseStudies')}
>
Case Studies
</Link>
03 Performance
By avoiding heavy DOM modifications, layout recalculations remain zeroed out during transitions. Page transitions are handled by Framer Motion's AnimatePresence.
The Three.js background is the most expensive element on the page. A particle system at full quality runs 2,000 instances at 60fps — fine on a MacBook, catastrophic on a mid-range Android. The solution is a device capability check on mount that drops particle count to 400 and halves the render resolution on low-end devices.
adaptation
04 Design System
The visual system is built around a single constraint: no color. The palette is six tokens — background, surface, border, text, muted, accent — all within a warm near-monochrome range. No brand color. No gradient. No shadow that isn't earned by the component that casts it.
“Simple can be harder than complex. You have to work hard to get your thinking clean to make it simple. But it’s worth it in the end because once you get there, you can move mountains.”
The automation work came from a practical rule: if a creative task repeats often enough, it should become a tool. The goal was not to replace taste. It was to compress the boring layer around editing, exporting, collecting, comparing, and batching so human attention could stay on selection and polish.
01 System Pipeline
The evidence matters because it shows workflow engineering around real production bottlenecks: thumbnail downloading, collage creation, frame extraction, beat-synced video generation, batch exports, and structured output logs.
01InputImages, clips, audio.
02AnalyzeFrame extraction, checks.
03GenerateMontages, batch cuts.
04QueueAsync subprocesses.
05ReviewHuman polish.
02 What Stayed Human
Automation did not decide what looked good. It made more options available faster. The human layer remained contrast, pacing, story hook, typography, emotional read, and knowing when an output was strong enough to publish.
Technical
FFmpeg and Python handle repeatable operations: slicing, muxing, resizing, batching, and rendering.
Creative
The final decision remains visual judgment: which frame, which rhythm, which composition deserves to ship.
These prototypes explore a practical question: how do multimodal models become actual product flows instead of chat demos? The answer is not just calling an API. It is image handling, caching, user state, deterministic recalculation, storage, canvas composition, and interface constraints.
01 Calories AI
Calories AI treats food recognition as a product architecture problem. Repeated vision calls can become slow and expensive, so the interesting layer is the LRU cache, perceptual hashing, and nutrition recalculation after the model response.
InputMeal image.Uploaded or captured food photo.
ModelVision analysis.Estimate foods/macros, structure the response.
LogicCache and recalculation.Reduce calls and update nutrition live.
02 Outfit Generator
The wardrobe tool connects uploaded clothing images, classification, storage, outfit composition, and generative recommendations. The product work is the bridge between messy personal assets and a clean visual output.
The Shopify work should not be presented as “I installed a theme.” The real work is platform constraint solving: understanding theme sections, snippets, product media, cart behavior, variant logic, editor schemas, and then inserting custom storytelling modules without breaking commerce reliability.
01 Shopify Constraints
For ADG Shop, the system had to support fast drops, product storytelling, creator-brand expectations, and the boring but critical mechanics of e-commerce: cart, variants, product media, checkout trust, and theme-editor stability.
Constraint
Shopify themes are opinionated. You cannot treat them like a blank React canvas.
Approach
Understand the theme architecture first: layout, sections, snippets, templates, assets.
Result
Custom sections and interactions can feel branded while still respecting the platform.
02 Rugzy Custom Srl
Rugzy needed more than product cards. A custom rug is only obvious when the transformation is shown: digital artwork to physical tufted texture, before/after, process, size, material, proof, and payment trust.
UniSpot turns an institutional portal into a student question: where can I sit right now? The work is not the UI alone; it is discovering how Unimore exposes room/building data, normalizing it, and presenting availability in a faster interface.
01 Data Problem
The university already had room data, but it was shaped for a portal, not for a student in motion. The product problem was to turn buildings, rooms, dates, events, and availability into a simple decision interface.
02 Interface Logic
01QueryBuilding and date parameters.
02NormalizeRooms, events, building codes.
03CacheSWR-style fetching.
04MapUser orientation.
05AnswerAvailable rooms shown.
Gstaad Guy / Utopia / MoreAmore
2026
Luxury visual systems · identity · apparel · motorsport concepts
This section separates projects that should not be mixed. Gstaad Guy, Utopia, MoreAmore, and Audi F1 x ADG are different visual problems with different audiences, materials, and levels of importance.
01 Gstaad Guy
Gstaad Guy is the strongest brand/visual proof in this group. The work is not "luxury moodboard." It is research-led visual direction: alpine luxury references, hotel-specific source material, social format analysis, quote/definition slides, first-slide variants, and final story exports.
02 Utopia & MoreAmore
Utopia Profumeria is a boutique identity problem: logo, packaging direction, business cards, stickers, ribbons, and a calm visual world for perfume retail in Modena. It should read as a compact brand identity, not be merged with unrelated apparel work.
MoreAmore is a separate apparel campaign: tee concepts, print preparation, DTF files, mockups, slogan systems, AI tests, feedback loops, and final production PDFs.
Amazon Prime Video Key Art
2025
Platform key art · delivery specs · safe-zone logic
The Amazon Prime Video work should not read as “I made a poster.” Streaming key art is a delivery system: the same artwork has to survive TV, mobile, category tiles, safe areas, cropping, title boundaries, face placement, export formats, and platform specs.
01 Platform Constraints
The production folder matters because it proves delivery discipline: PSD/PSB source files, PNG exports, PDF recaps, generated image iterations, RAW/JPG material, font packages, and promotional material delivery specifications.
Safe Area Thinking
Artwork that looks correct in Photoshop can fail inside a streaming UI. The design has to account for dynamic crops, title safety, focal points, and device-specific framing.
02 What I Delivered
VisualKey art production.Composition, title placement, exports.
ProfessionalOfficial media context.Production quality under platform constraints.
Nightmare GP: Motorsport Concept
2026
Motor Valley Fest 2026 · Innovation Award · concept system
Nightmare GP should never be reduced to telemetry overlays. The real work was turning a hackathon concept into a coherent motorsport product world: format design, brand language, pitch logic, web experience, car/circuit assets, deck structure, motion, and presentation.
01 Format Design
The concept was designed around entertainment logic: compact night racing, Gen Z pacing, short technical tracks, team identity, elimination structure, and recognizable mechanics like shortcuts, boost, and Nightmare Mode. The point was not a gimmick; it was a format that could be explained, watched, branded, and pitched quickly.
Format16 drivers, 8 teams.A clear structure for short-form motorsport.
Brand worldNight racing, LED identity.A visual language that makes the format ownable.
RecognitionInnovation Award.Presented in an automotive innovation context.
02 Site & Pitch System
The site and deck made the concept feel real. A pitch can describe an idea; a visual system makes it believable. The evidence includes solution documents, deck versions, videos, circuit assets, screenshots, and a React/Vite/Three-style web layer.
Aiman's Finance OS
2026
Regime Forfettario · income dashboard · personal operating tool
Aiman's Finance OS is a small but useful product direction: a dashboard for translating Italian freelance income into more understandable operating numbers. It belongs in the portfolio as a tool/product example, not as a main headline.
01 Live Calculator
The useful part is not the dashboard surface alone. It is turning gross freelance income into an immediate operating number: taxable base, 5% substitute tax, contribution estimate, total deductions, and net income.
Realtime estimator
Gross income → estimated net income
78% base · 5% tax · 27.72% INPS
EUR / year
€0Estimated net income calculated live.
Taxable base€0
5% tax€0
Contributions€0
Total deductions€0
calc.ts
// Live calculation model
const base = grossIncome * 0.78;
const taxes = base * 0.05;
const contributions = base * 0.2772;
const netIncome = grossIncome - (taxes + contributions);
02 Product Logic
The point is operational clarity: instead of keeping the tax model as an abstract formula in a note, the dashboard makes the number feel live. Changing revenue immediately changes the taxable base, deductions, and net projection.
This is a simplified estimator for portfolio/product demonstration, not fiscal advice. The rates and assumptions are intentionally visible so the model can be inspected and adjusted.
Non so programmare. Ho costruito lo stesso.
June 2026
On taste, tools, and what actually changed with AI
The contradiction
There is a contradiction at the center of my work that I stopped hiding. I am the founder of a creative studio, I direct complex digital projects, I have built 3D configurators, Shopify stores, visual systems for real clients — and I have never written a line of code by hand.
At first this felt like something to apologize for. Then I understood it was the opposite.
The shift
For years the market rewarded people who could do everything: design and code, copy and strategy, creativity and analytics. The ideal profile was the full-stack creative — one person capable of covering the entire production cycle.
AI made this profile nearly irrelevant. Not because technical skills stopped mattering — they still do — but because the bottleneck moved. Anyone can now generate working code, decent layouts, mediocre copy. Scarcity is no longer in execution. It is in judgment.
The question is no longer "can you do it?" but "do you know when it's done well?"
I have used Claude, Cursor, Windsurf, and a dozen other tools to build things that five years ago would have required a team. It is not magic — it requires a different kind of intelligence: writing precise briefs, evaluating output, iterating with intention.
The problem is most people use these tools to produce more. I use them to produce better. The difference is subtle but total.
Taste as leverage
Taste is not easily taught. It accumulates through exposure: years of watching what works and what does not, understanding why certain visual choices communicate authority while others do not, developing an almost physical allergy to the mediocre.
When a client like ADG — 278 million lifetime views — asks you to direct the visual identity of a project, they are not paying for your hands. They are paying for your ability to see the problem in a way they cannot.
The moment
It was during the Nightmare GP pitch at Motor Valley Fest, in front of a jury that included Ferrari, Dallara, and Pirelli. We were not the most technical team in the room. But we had the clearest narrative, the most coherent design system, the best-constructed story.
We won the Innovation Award.
Modena is not Silicon Valley. But it has something the Valley does not: a visceral relationship with material excellence. Ferrari, Maserati, Lamborghini, Ducati — these companies did not build legends through feature sheets. They built them through emotion engineered with millimetric precision.
That is the approach I want to bring to digital work. Not faster. Better.