Loading

JT— back to top

Student • AI Enthusiast • Developer

Jubayer Tahsin

LearnBuildInnovateLead

Exploring Computer Science, building with AI, and preparing to create technology that matters.

View projectsGet in touch

Bangladesh

Portfolio 2026

01 — About

I am an HSC Science student from Bangladesh with strong interests in Computer Science, Artificial Intelligence, Software Engineering, Cybersecurity, and Technology Entrepreneurship.

Reading about a system tells me how it is supposed to work. Building one tells me how it actually does. So most of what I know came out of practical experimentation, real projects, technical problems that refused to resolve, and a lot of troubleshooting.

Three of those interests are usually kept in three separate portfolios. To me they are one job. A model nobody can audit is a model nobody should ship, and neither of those matters if the thing has no surface a person can actually use.

Status
HSC Science student
Based
Bangladesh
Studying
Science, headed for Computer Science
Graduating
2026

How

  • Practical experimentation
  • Building real projects
  • Solving technical problems
  • Exploring future-oriented technologies
  • Troubleshooting
  • Hands-on development

Learn. Build. Innovate. Lead.

02 — Skills

No percentages and no levels, because there is no honest way to put a number on any of this. 8 groups, and every line between them is drawn from something they actually share.

  • The layer underneath everything else here.

    • Programming
    • Software Engineering
    • Problem solving
    • Troubleshooting
    • Automation

  • What I write in, and what I write with.

    • Python
    • JavaScript
    • HTML
    • CSS
    • Git
    • GitHub
    • API Integration

    Used in POGO, FreeLLMAPI, OmniRoute, VQT

  • Not only using models. Building the systems around them.

    • LLMs
    • Multi-model AI
    • AI Agents
    • AI Memory
    • Embeddings
    • Model Routing
    • Local AI
    • Cloud AI
    • Tool Use
    • AI APIs
    • AI Automation
    • Machine Learning
    • Personal AI
    • AI Operating System

    Used in POGO, Pakhi AI, FreeLLMAPI, OmniRoute

    Further AI projects on GitHub, not yet written up here.

  • The part people actually touch.

    • Web Development
    • Static Websites
    • HTML
    • CSS
    • JavaScript

    Used in Pakhi AI

    Web projects on GitHub, not yet written up here.

  • The reason I try to break what I build.

    • Cybersecurity

    Cybersecurity projects on GitHub, not yet written up here.

  • Where the software has to actually live.

    • Linux
    • Networking

    Linux and networking projects on GitHub, not yet written up here.

  • Making a thing run somewhere other than my machine.

    • Docker
    • Dockerized Applications
    • Self-hosted Applications
    • Cloud Infrastructure
    • DevOps

    Used in Pakhi AI

    Docker projects on GitHub, not yet written up here.

  • The half of the work that is about attention.

    • Video Editing
    • Creative Content
    • Reels
    • YouTube

Eight groups, connected where they genuinely overlap. Hover or focus one to see what it shares and with which project.

The AI stack I keep coming back to

  • ChatGPT
  • OpenAI
  • Claude
  • Gemini
  • Gemma
  • OpenRouter
  • FreeLLMAPI
  • OmniRoute
  • LM Studio
  • Ollama
  • Open WebUI
  • OpenClaw

The shape it keeps taking

  1. 01Local AI
  2. 02Cloud AI
  3. 03Model Router
  4. 04Memory
  5. 05Tools
  6. 06Automation

03 — Work

Infrastructure more than apps: the layer underneath the model rather than the prompt on top of it. Every status below is the real one.

01

POGO

Personal AI Operating System / Personal AI Assistant

Developing

A personal AI operating system: many models, a memory that persists, tools it can actually use, and automation on top. One assistant that knows the whole context instead of six chat tabs that each know a sliver.

The problem

Every model lives in its own window with its own blank memory. Nothing knows what I did yesterday, which model is right for this question, or how to reach my files.

The approach

Treat it as an operating system rather than a chat app. Route between local and cloud models, keep one persistent memory underneath them, give the whole thing tools, and let automation drive it.

Built with
  • OpenClaw
  • LLM APIs
  • Local AI
  • Cloud AI
  • Memory
  • Automation
Ideas
  • Model routing
  • Persistent memory
  • Tool use
  • Multi-model AI
  • Personal AI

What it is meant to do

  • Personal memory
  • Multiple AI models
  • Model routing
  • Tool use
  • Automation
  • Coding assistance
  • File interaction
  • Local AI
  • Cloud AI
  • Project assistance

Earlier in development the agent went by Crestodian.

Repository not published yet

02

Pakhi AI

AI Interface / Open WebUI Customization

Developed / Experimental

A personal AI interface built by customizing Open WebUI, so talking to local and hosted models happens somewhere that feels like mine rather than somewhere generic.

The problem

A default interface shapes how you use a model. I wanted the environment to fit the way I actually work with AI.

The approach

Start from Open WebUI rather than from nothing, and customize the surface into a personal environment for local and cloud models.

Built with
  • Open WebUI
  • Local AI
  • Self-hosted
Ideas
  • AI interface
  • Local AI
  • Personal environment

Experimental by design. The interface is the experiment.

Repository not published yet

03

FreeLLMAPI

AI Infrastructure / LLM API

Completed / Running

One interface in front of many language models and providers, so an application can ask a question without first learning six different API shapes.

The problem

Every provider has its own request format, its own auth, its own quirks. Supporting four of them means writing the same integration four times.

The approach

Abstract the model behind a common interface, and let the platform absorb the differences between providers.

Built with
  • LLM APIs
  • Multi-provider AI
Ideas
  • Unified interface
  • Model abstraction
  • Multi-provider AI
  • AI infrastructure

Repository not published yet

04

OmniRoute

AI Model Routing / Infrastructure

Completed

A routing layer between an application and many models, so switching provider is a configuration change rather than a rewrite.

The problem

Building against one provider's API is building a dependency on that provider. Swapping models later means touching everything.

The approach

Put a router in the middle. The application talks to the router, the router talks to whichever provider fits, and the application stays provider-independent.

Built with
  • LLM APIs
  • Multi-provider AI
  • Model routing
Ideas
  • AI routing
  • Provider independence
  • Usage optimization
  • AI infrastructure

Repository not published yet

05

VQTVideo QR Data Transfer

Data Transfer / Experimental Technology

Developing

Moving data between devices as pictures. Encode the payload into QR frames, play them as video, and let the other device read it back with a camera.

The problem

Two devices that cannot be paired, networked, or trusted with a cable still have a screen and a camera between them.

The approach

Use visual encoding as the transport: frames of QR carried by video, decoded on the far side.

Built with
  • QR
  • Video
  • Visual encoding
Ideas
  • Experimental technology
  • Visual encoding
  • Data transfer

An experiment first. Whether it is practical is part of what is being tested.

Repository not published yet

The one it is all pointing at

Personal AI Operating System

Developing

Local AI
Runs on my own machine. Private, offline, no per-token cost.
Cloud AI
Reached for when the question is bigger than the local model.
Model Router
Decides which model answers, so the rest of the system never has to care.
Memory
What makes it mine rather than a fresh assistant every morning.
Tools
Files, code and the projects I am actually working on.
Automation
The work that should happen without me asking for it twice.

Earlier in development the agent went by Crestodian.

04 — Direction

Build strong technical skills, create real software and AI products, generate income through technology, and eventually build and lead a technology-driven company.

  1. Now

    HSC Science, and every spare hour spent building.

    Now

  2. Learning

    Computer Science properly, at degree depth.

    Now

  3. Building

    Real software and AI products, shipped and used.

    Ahead

  4. Innovating

    Systems around AI, not just applications on top of it.

    Ahead

  5. Leading

    A technology-driven company of my own.

    Ahead

The long gameCEO

Long-term aspiration, not a current title.

05 — Experience

No formal employment yet. This is the work itself, which is the part that taught me anything.

  • 01

    AI systems

    Four projects that are infrastructure rather than apps.

    POGO, FreeLLMAPI, OmniRoute and Pakhi AI all exist because I wanted to understand the layer underneath the model, not just the prompt on top of it.

  • 02

    Experiments

    Ideas built far enough to find out whether they work.

    VQT is the clearest example: an experiment in moving data as video, built to be tested rather than to be right.

  • 03

    Web development

    Static sites and interfaces, written by hand.

    HTML, CSS and JavaScript, plus the interface work in Pakhi AI. The surface is not the easy half.

  • 04

    Video editing

    YouTube video editing for two teachers.

    Real edited output on a real schedule, which is a different discipline from writing code and made me better at pacing anything.

06 — Education

  1. Current

    HSCScience

    Barguna Govt. College

    Board
    Barishal Education Board
    Session
    2024 - 2025
    Status
    Graduating 2026

    HSC candidate. Science stream, and the reason the rest of this list exists.

  2. Intended

    Bachelor's degree

    Status
    Next
    Preferred
    Germany

    The plan is a bachelor's degree in one of four fields, with Germany as the preferred destination. Nothing is decided or applied for yet, and this section will say so until it is.

    One of four

    • Computer Science
    • Computer Science & Engineering
    • Software Engineering
    • Cybersecurity

07 — Capability

Not an agency. One student who builds these things.

  • 01Web DevelopmentSites and interfaces, written rather than assembled.
  • 02AI DevelopmentApplications built on language models.
  • 03AI AutomationWork that should not need a human twice.
  • 04Software DevelopmentTools, scripts and small systems.
  • 05AI API IntegrationWiring models into something that already exists.

08 — Proof

The proof is in the repositories

Named projects above. Everything else, including the work that has not been written up, is here.

github.com/xubayertahxin (opens in a new tab)

What is in there

  • AI Projects
  • Web Projects
  • Cybersecurity Projects
  • Networking Projects
  • Docker Projects
  • Linux Projects