av.Adrian Versoza
Let’s talk

Automation / Integrations / AI

Connect systems.
Make work flow.

Hi, I’m Adrian. I connect services, automate workflows, and build AI-powered tools with Python, APIs, and a hands-on approach.

Explore my workLet’s talk
Prototype. Test. Document. Improve.
FROM IDEA TO EXPERIENCE01 / 02
Web. Bots. AI.
Built from real problems.
Akbash / moderation
Akbash staged product capture showing AutoMod controls
Decision Diary home screen with upcoming reviews and recent decisions
BUILT WITH INTENTION
Two projects. One curious mind.
APIs & integrationsPython automationAI-powered tools

01 / Selected work

Proof in the projects.

Connecting systems. Handling edge cases.
Building beyond the demo.

CONNECTED SYSTEMSWEB + BOT

Sanitized product demo captures

Akbash / AutoMod

Sanitized Akbash product demo showing global and channel-specific AutoMod controls

Sanitized product demo captures

Web platform / Discord automation01

Akbash

A moderation workflow that connects Discord, a Python bot, a web dashboard, and AI assistance. Rules, cases, appeals, and support become one coordinated system.

PythonNext.jsDiscord APIAI APIs

APIs, automated actions, and reliable recovery.

Inside the build
01

One shared source of truth

PostgreSQL keeps the bot and dashboard in sync, with notifications and Redis coordinating work across services.

02

Actions that can recover

Transactional queue claims, leases, and retries recover interrupted work. Discord rate limits trigger cooldowns, while permission denials stop retries.

03

Permissions at the boundary

Discord OAuth, server-scoped guards, and permission rechecks keep moderation actions tied to the right authority.

04

Model APIs with a human checkpoint

A streaming gateway integrates Anthropic, Gemini, and OpenAI-compatible APIs, including local Ollama. Proposed server rule changes require confirmation.

05

Documentation for the next person

Workflow notes explain service signals, sequence diagrams, recovery behavior, and migration rollback so the system can be maintained.

Complete project stack

Next.jsReactTypeScriptTailwind CSSPythondiscord.pyDiscord APIAnthropic APIGemini APIOpenAI-compatible APIsOllamaPostgreSQLPrismaasyncpgRedisNextAuthZodDocker ComposePgBouncerNGINX

Regression tests cover provider networking, permission checks, and rejected background actions. Tooling: Vitest, pytest, ESLint, and Ruff.

Visit Akbash
A PRIVATE SPACEANDROID

Sample journal entries from the product

Decision Diary / Home

Decision Diary home screen with sample entries and review dates

Sample journal entries from the product

Native Android / On-device AI02

Decision Diary

A private decision journal for recording your reasoning, revisiting outcomes, and reflecting with an AI companion that runs entirely on your device.

KotlinJetpack ComposeRoomLiteRT LM

Local AI, bounded resources, and useful fallbacks.

Inside the build
01

Privacy by construction

The app has no Internet permission. Journal data and AI inference stay on the device.

02

Local AI with a fallback

Qwen3 runs through LiteRT LM in a separate inference process, with bounded context and a basic reflection fallback.

03

Persistence without shortcuts

Room transactions and explicit migrations preserve decisions, reviews, and conversations as the app evolves.

04

Respect the device

Background review reminders use WorkManager, while native AI resources are released when a conversation closes.

Complete project stack

KotlinJetpack ComposeMaterial 3RoomCoroutinesStateFlowWorkManagerLiteRT LMQwen3 0.6B INT4Android KeystoreBiometricPromptJUnitRobolectric

Validation tooling: JUnit, Robolectric, Android instrumentation, and Baseline Profiles.

Explore Decision Diary

02 / How I build

A problem to solve.
A tool to build.

I’m Adrian John E. Versoza. I use code and AI tools to build websites, applications, and Telegram and Discord chatbots. Python is my preferred way to automate, with n8n’s AI tools also part of my experimentation.

I use Git and GitHub for project development, tracking changes, and version control. I also work with Docker and Kubernetes for containerized applications.

I’m looking for a hands-on role where I can work closely with founders, turn loosely defined problems into working prototypes, and own the path from testing to documentation and handoff.

See what I’m building on GitHub
01

Connect the systems.

Discord APIs, a web dashboard, shared data, and AI providers come together in Akbash.

02

Code the missing piece.

Python services handle moderation actions and recovery. TypeScript connects the dashboard to the workflow.

03

Test beyond the demo.

Permissions, retries, data migrations, and AI fallback paths are part of the build, with tests and project notes to support them.

Tools I work with

A stack chosen for the problem.

APIs & integrations

Discord APIDiscord OAuthAI provider APIsNext.jsTypeScript

Android & local data

KotlinJetpack ComposeRoomCoroutines

Code & automation

Pythonn8nPostgreSQLRedis

Development workflow

GitGitHubVersion controlProject development

Containers & delivery

DockerDocker ComposeKubernetesNGINX

AI & experiments

LiteRT LMQwen3Google Colabn8nPython automationGeminiChatGPTClaude

Testing & handoff

VitestpytestJUnitRobolectricESLintRuffArchitecture notesSetup & recovery guides

From a problem to a working tool

How I approach the work.

  1. 01

    Scope

    Understand the workflow, users, constraints, and what a useful first version needs to do.

  2. 02

    Prototype

    Connect the core pieces using code, APIs, and AI tools. Make the idea concrete early.

  3. 03

    Test & iterate

    Try failure cases, inspect the results, and use feedback to decide what to improve next.

  4. 04

    Document & hand off

    Explain setup, behavior, recovery, and limits so the next person can maintain the tool.

03 / The AI lab

AN ONGOING EXPLORATION

Stay curious.
Keep experimenting.

Building products teaches me how systems work. Experimenting with AI gives me new ways to think about what they can do.

Small experiments. Practical questions.
01 / Google ColabFine tuning

Small models, focused tasks.

I use Google Colab to experiment with fine tuning small models and explore how their behavior can be adapted.

02 / n8nAI workflows

Connect the pieces.

I’ve used n8n, mainly its AI tools, to explore how model capabilities can fit into an automated workflow.

03 / PythonCoded automation

Code the workflow.

Python is my preferred approach to automation. I like being able to define the logic and control how the steps work together.

04 / Gemini, ChatGPT & ClaudeAI tools

Build with AI. Understand the system.

I use AI tools to help create websites, applications, and Telegram and Discord chatbots, then work through how the pieces fit and behave.

Fine tuning and automation are areas of experimentation. On-device AI is implemented in Decision Diary.

04 / Make a connection

A workflow
worth fixing?

Let’s talk about automations, AI tools,
or a hands-on builder role.

Email
Phone+63 981 254 0199
ElsewhereGitHub / adrianjohn141