Sean's Portfolio

My Skills

Programming Languages

JavaScript

Python

Java

Web Technologies

HTML/CSS

React

Node.js

AI & Machine Learning

TensorFlow

NLP

Reinforcement Learning

About PEARL

PEARL is an experimental, self-improving AI assistant currently under active development. Designed as the core agent within SeanOS, PEARL moves beyond conventional assistants by integrating dynamic function execution, contextual memory, goal planning, and self-editing capabilities.

Rather than operating through rigid command trees, PEARL interprets user intent using local large language models (LLMs), autonomously selects and executes appropriate functions, and continuously reflects on its own performance to improve over time. It features a modular architecture that allows it to interact through multiple interfaces, including Telegram and a web-based UI, while maintaining persistent memory across sessions.

Key Features

Dynamic intent parsing and function execution
Persistent, topic-based memory with daily resets and summaries
Self-editing and self-reflection modules for autonomous improvement
Integration with APIs, external tools, and multimodal inputs (text, images, video)
Knowledge expansion and real-time research capabilities with fact validation

PEARL is designed not just to assist, but to evolve—growing increasingly aligned with the user's goals through continuous adaptation, memory consolidation, and self-directed learning. It serves as the foundation for exploring lifelong personal AI and the early steps toward general agentic intelligence.

About SeanOS

SeanOS is an experimental cognitive operating system designed to serve as a lifelong, autonomous digital companion. Currently under active development, SeanOS integrates memory, reflection, modular intelligence, and multimodal interaction to create a system that can reason, act, and evolve alongside its user.

Unlike conventional operating systems, SeanOS is not limited to managing computational resources—it is built to manage cognitive processes. It enables persistent goal tracking, autonomous task execution, episodic memory storage, and dynamic learning through an integrated AI agent: PEARL.

Core Concepts

Memory-Enabled Maintains long-term context via Git-based episodic memory, allowing it to reflect on prior interactions and refine behavior over time.
Goal-Driven Operates based on user-defined and self-inferred goals, adapting its behavior and actions accordingly.
Modular & Extensible Supports dynamic loading of executable Python modules, including self-generated code.
Multi-Interface Interacts across platforms such as Telegram, CLI, and a web-based interface modeled after ChatGPT.
Self-Improving Periodically performs self-reflection, documentation updates, and architectural optimization.

SeanOS is an attempt to reimagine what operating systems can become in the age of artificial intelligence—not just software that runs programs, but a system that understands, remembers, and collaborates with its user across the arc of their life.

✨ Lumina - Privacy-First Personal Memory Assistant

Lumina is a privacy-first personal memory assistant that helps you remember and recall information from your documents. It combines semantic search, intelligent recall, and conversational AI to create a powerful personal knowledge management system.

🔧 Key Features

Document Ingestion: Upload PDFs and text files for processing and indexing
Semantic Search: Find relevant information using natural language queries
Intelligent Recall: Get contextual answers based on your stored knowledge
🤖 Chat with LLM: ChatGPT/Gemini-like conversational interface with document augmentation
💬 Persistent Chat Sessions: Maintain conversation history across sessions
Privacy-First: Local processing with optional cloud deployment
Local LLM Support: Powered by Ollama for complete privacy
Beautiful Web Interface: Streamlit-based UI for easy interaction

Architecture & Technology

FastAPI Backend High-performance API server handling document processing, search, and chat functionality
Streamlit Frontend Intuitive web interface for document upload, search, and conversational AI
ChromaDB Vector Store Efficient semantic search and document retrieval using vector embeddings
Ollama Integration Local large language model processing for complete privacy and control
Modular Design Extensible architecture supporting multiple document types and LLM providers

Use Cases & Applications

Personal Knowledge Management: Organize and search through personal documents, notes, and research
Academic Research: Process and query academic papers, textbooks, and study materials
Business Intelligence: Analyze company documents, reports, and knowledge bases
Content Creation: Research and reference materials for writing and content development
Learning & Education: Interactive study assistant with document-based Q&A

Lumina represents a shift toward truly personal AI systems that respect privacy while providing powerful knowledge management capabilities. By keeping processing local and maintaining user control over data, it offers a secure alternative to cloud-based AI assistants while delivering sophisticated document understanding and conversational capabilities.

The Thinking Room

A collection of writings, essays, and research papers exploring the deeper questions of artificial intelligence, consciousness, and the nature of intelligence itself.

📄 The Illusion of Generality

Why Artificial General Intelligence May Be Fundamentally Impossible

Read PDF

About Me

Hi, I'm Sean — a computer science undergraduate exploring the limits of artificial intelligence, particularly the question of whether true artificial general intelligence (AGI) is even achievable.

My work focuses on building autonomous, goal-driven systems that can reason, reflect, and adapt — not because I believe these will evolve into full AGI, but because they offer valuable insight into cognition, control, and the boundaries of machine intelligence.

I'm currently developing three ongoing projects:

PEARL — a self-improving AI assistant designed for dynamic function execution, contextual memory, and multimodal reasoning.
SeanOS — a cognitive operating system that integrates long-term memory, goal management, and reflective processing to serve as a lifelong personal AI framework.
Lumina — a privacy-first personal memory assistant with document ingestion, semantic search, and conversational reasoning.

Together, these projects investigate how far modular architectures, dynamic planning, and memory-driven behavior can take us — and where the true limits of "intelligence" may lie.

My broader interests include explainable AI, reinforcement learning, personal informatics, and human-AI collaboration. I approach AGI not as an inevitability, but as a philosophical and technical frontier worth questioning.

Let's Work Together

Get In Touch

✉️

Email

tan.jun.kit.sean@gmail.com

📍

Location

Singapore

Connect With Me

The best way to reach me is directly through email. I look forward to discussing potential collaborations and opportunities in AI research and development.

For professional inquiries, you can also connect with me on my social media profiles linked above.

Hello, I'm Sean