Aislam Camarini

AI Product Engineer · Creative Technologist

I build digital products, AI systems, agents and interactive experiences — from concept to deployment.

I build ideas that live between

technology, design and intelligence.

scroll to cross over

Manifesto

I don't work within a single discipline. I build connections between them.

Engineering, AI, product and creative direction in the same process. Every project starts with a real problem and ends as a system that works — and that makes sense at first glance.

Selected work

Five chapters, five worlds.

012026Open source · alpha

IsaHat

Open-source security auditor for modern web applications, APIs and systems.

Problem
AI-assisted applications ship fast and repeat the same weaknesses: incomplete authentication, broken authorization, exposed secrets, fragile APIs.
System
One Python audit engine shared by a CLI, a desktop app (Tauri + React) and a local bridge API. Every finding carries evidence, a confidence level, impact, a fix and how to verify it. Safe by default, offline-first.
My role
Designed and engineered end-to-end

Python · CLI · Tauri + React · SQLite · CI/CD · Security · Developer Tool

022026Open source · pre-alpha

Noema

Open-source adaptive learning platform that models what you understand — and what you're about to forget.

Problem
Study tools store what you put in. None of them know which concepts you actually master, or why you keep getting something wrong.
System
Next.js + FastAPI, PostgreSQL with pgvector, Redis and background workers. Extracts concepts from your material, builds a knowledge graph, schedules reviews with FSRS and answers only from your sources, with citations. Runs fully local with Ollama.
My role
Designed and engineered end-to-end

Next.js · FastAPI · PostgreSQL + pgvector · Redis · RAG · Adaptive Learning · EdTech

032026Open source · alpha (research)

Sabelia

The neural network that acts as Noema's brain: it learns, from study events, what each person knows, what they are forgetting and what they should do next.

Problem
A language model writes well, but it does not know what a learner has mastered or when they will forget it. A learning product needs a brain of its own to decide the next action.
System
A neural sequence model (causal attention with a forgetting gate) trained on study events, evaluated with calibration and versioned in a registry. A pedagogical policy returns the next action with reason codes, through FastAPI. The LLM stays outside: Noema asks Sabelia what the learner needs. On synthetic benchmarks the baselines still win — and the README says so.
My role
Designed and engineered end-to-end

Python · PyTorch · NumPy · FastAPI · Model Evaluation · Neural Network · Machine Learning · Research

042026Open source

ISLA

Open-source virality radar: signals in, trends out — measured between real samples, never invented.

Problem
Trend tools show numbers with no origin. Spotting what is about to go viral means measuring acceleration between samples, not size.
System
A Python engine that collects public sources (Google Trends, Wikipedia, Bluesky, Hacker News, RSS), computes velocity and acceleration, and exposes REST + SSE for a live control room. Provenance for every number; optional clustering through the Anthropic API.
My role
Designed and engineered end-to-end

Python · asyncio · SQLite · REST + SSE · Anthropic API · Data Systems · Viral Intelligence

052026Ongoing

Brand Systems

Identities and visual systems for digital brands.

Problem
A digital brand needs a system, not just a logo: typography, color, motion and rules that outlive whoever applies them.
System
Creative direction and systems design: identity, typography, visual components and motion, documented for consistent use.
My role
Creative direction and design

Identity · Typography · Design Systems · Motion · Branding · Art Direction

What I build

From product to system, end to end.

  • Agents, RAG, model orchestration and evaluation — products where the LLM is one part of the system, not the whole system.

About

Technology with creative thinking. Creativity with structure.

I'm Aislam Camarini, a Brazilian AI Product Engineer and Creative Technologist. I work at the intersection of engineering, AI, product and design, turning ambitious ideas into working systems and well-built digital experiences.

Self-taught, I build products end-to-end — from the data model to the interface — and keep part of my work open source.

Location
Santos, Brasil
Languages
  • PortuguêsNative
  • EnglishFluent
  • EspañolIntermediate
  • ItalianoIntermediate
Availability
  • Available for international projects
  • Remote work
  • Open source
  • Collaborations
  • Experimental projects

Skills

A system of skills, not a list.

AI Engineering

  • LLMs
  • AI Agents
  • RAG
  • Model Evaluation
  • Prompt Engineering
  • Machine Learning
  • PyTorch

Software Engineering

  • Python
  • TypeScript
  • JavaScript
  • React
  • Next.js
  • FastAPI
  • PostgreSQL
  • REST APIs
  • Git

Creative Technology

  • Three.js
  • WebGL
  • GSAP
  • Interactive Design
  • Generative Systems

Product & Design

  • Product Engineering
  • UX/UI
  • Design Systems
  • Branding
  • Creative Direction

Experiments

Laboratory

Every study here is running on this page: these are the techniques behind what you just scrolled through.

  • GLSL · WebGL

    Terrain Shader

    Procedural terrain with simplex noise and scroll-driven palette.

    in the contact valley

  • Canvas · Generative

    Halftone Field

    Cursor-reactive dot field with a halftone weave.

    in the “What I build” organism

  • Motion · Typography

    Type Reveal

    Line splitting and sequential reveal of the headings.

    in every heading on this page

  • Identity · Generative

    Glyph System

    Modular glyph system derived from the A/M symbol.

    in the brand symbol

Contact

Have an interesting problem? Let's talk.

Available for ambitious products, collaborations and international opportunities.

contact@aislam.dev

© 2026 Aislam Camarini Mastro

Santos, Brasil · built with Next.js, GSAP and Three.js