Rahadian Bayu Permadi

Rahadian Bayu Permadi

Mentor
Rising Codementor
US$17.00
For every 15 mins
ABOUT ME

Software architect and full-stack engineer with 15+ years across the US, Europe and Asia — most recently two years at Woven by Toyota, where I built a production AI assistant (RAG over internal documentation on AWS Bedrock with Claude and Llama) and refactored a codebase of 10,000-line files into modules under 500 lines, taking test coverage from about 15% to about 80%.

I build AI into products that ship, on backends I build myself in Go, Python and TypeScript on AWS. Before Woven I took a social marketplace app from paper design to tens of thousands of users.

I’m a product-minded engineer — I look past the ticket to whether the solution actually delivers user value and keeps the system stable. I also maintain perso, an open-source authorization engine for AI agent tool calls.

Hours: 9am–7pm UTC+7 — full overlap with Singapore, Japan and Australia, and mornings with Europe. I also work evenings for US Eastern clients: 9pm Jakarta is 9am New York.

Jakarta (+07:00)
Joined June 2020
EXPERTISE
10 years experience
9 years experience
7 years experience
6 years experience
9 years experience
RAG
2 years experience
2 years experience

REVIEWS FROM CLIENTS

Rahadian's profile has been carefully vetted and approved as a Codementor. Connect with Rahadian now, and leave a review for them once you're done!
SOCIAL PRESENCE
GitHub
flask-graphql-template
Flask project template with user login and everything basics
Python
3
1
perso
a WebAssembly-compiled policy enforcement engine for MCP (Model Context Protocol) tool calls. You define access rules in a JSON file, compile them into a portable .wasm binary, and embed that binary in any host (Rust, Node.js, Python, Go, etc.) to make Allow/Deny decisions on every LLM tool call before it reaches MCP.
Rust
1
0
EMPLOYMENTS
Platform Software Engineer
Woven by Toyota
2024-09-01-2026-09-01
  • Built a production AI assistant: Embedded an LLM chat assistant in the internal developer portal, using retrieva...
  • Built a production AI assistant: Embedded an LLM chat assistant in the internal developer portal, using retrieval (RAG) over the platform’s own documentation with Amazon Bedrock (Claude and Llama), so engineers get answers in a single turn instead of searching through docs. A second layer of MCP-based actions — creating users and groups from within the conversation — was built and pending release.
  • Replatformed the internal developer portal: Led the migration from Backstage to a custom Next.js application in TypeScript and React with Tailwind CSS, working from Figma designs, centralizing infrastructure discovery and service management for global teams.
  • Cross-functional technical leadership: Worked with engineering leads across regions to translate business requirements into technical blueprints, meeting the security and compliance standards of the Toyota ecosystem.
  • Improved code quality and testability: Refactored more than ten files of over 1,000 lines — several above 5,000, one close to 10,000 — into modules under 500 lines each, reducing complexity and raising test coverage from about 15% to around 80%.
  • Improved the development process: Broke work and JIRA tickets into small, testable units, scrutinized requirements to extract concrete actions for each ticket description, and introduced step-by-step test instructions for QA. The result was smoother delivery and faster onboarding for new team members.
  • CI/CD and infrastructure: Helped build GitHub Actions workflows (PR checks and tests before merge; deploys to dev, staging, production) on Terraform-managed blue-green AWS infrastructure.
Python
TypeScript
React
View more
Python
TypeScript
React
Terraform
Next.js
RESTful API
GitHub Actions
Fastapi
AWS
LLM
RAG
Backstage
Aws (eks, bedrock)
View more
Principal Full-stack Software Engineer
FuzeMee
2020-06-01-2023-09-01
  • From paper design to launch: Took a college social and marketplace app from a paper design to a live product wit...
  • From paper design to launch: Took a college social and marketplace app from a paper design to a live product with tens of thousands of users, launched in 2021 on both the App Store and Google Play — including certificates, provisioning, store listings and release cycles.
  • Cloud architecture: Designed and ran the full AWS architecture — Amplify, AppSync (GraphQL), DynamoDB, Lambda, S3 and Cognito — for a scalable, secure product, maintaining 99.9% availability.
  • Marketplace payments: Built checkout and seller payouts with Stripe.
  • Search and background processing: Built text and location-proximity search with OpenSearch, background jobs with SQS, and push notifications for iOS and Android.
  • Offline media sync: Built local persistence with backend sync for photo and video uploads, compressing and reducing quality when files were too large for the connection.
  • Team leadership: Led a cross-functional team and managed two developers in the US and China, running code reviews and architectural guidance, and turned US stakeholders’ business goals into technical roadmaps.
Android
Node.js
Elasticsearch
View more
Android
Node.js
Elasticsearch
TypeScript
Stripe
React Native
GraphQL
AWS DynamoDB
AWS Lambda
iOS
AWS Amplify
AWS
View more
Full Stack Developer
Syngenta
2019-12-01-2020-05-01
  • Internal developer platform: Built an enterprise platform that let scientists provision cloud services across AW...
  • Internal developer platform: Built an enterprise platform that let scientists provision cloud services across AWS, GCP and Azure to run their experiments — Go backend, React and Gatsby front end, deployed on OpenShift.
  • Cloud provisioning services: Wrote the Go services that create and manage cloud resources across all three providers.
  • Testability refactor: Led a refactor of the Go backend for unit testing and package modularity, taking test coverage from 0% to 30% and establishing the TDD practices the team carried forward. Production regressions dropped.
  • Engineering standards: Set the patterns developers followed when building new features, services and products.
PostgreSQL
Azure
Test driven development
View more
PostgreSQL
Azure
Test driven development
React
Google Cloud Platform
OpenShift
Microservices
Gatsby
Golang
AWS
View more
PROJECTS
Perso - an embedded ABAC policy engine for MCP tool-call authorizationView Project
2026
perso is an embedded ABAC policy engine for MCP tool-call authorization — compiled to WebAssembly, with no control plane, no SaaS account...
perso is an embedded ABAC policy engine for MCP tool-call authorization — compiled to WebAssembly, with no control plane, no SaaS account, and no network call in the decision path. You define who can call which tools, under what conditions, in a plain JSON file. That file compiles into a single portable .wasm binary you load directly into your own process — a backend server, an MCP server, an edge function, or a CLI. There is nothing to deploy, register, or authenticate against. The policy engine ships inside your binary, next to your code, and answers Allow/Deny in-process, in microseconds. The LLM never touches auth. The host owns the role. perso makes the call at the point where the tool call would be forwarded — without leaving the process. perso is stateless and per-call by design: each decision evaluates one tool call against the attributes you pass in. For agentic loops — where the real risk often lives in a sequence of individually-innocent calls — the host or gateway accumulates session context and feeds it in as attributes; perso evaluates whatever context you give it, at microsecond cost per call. This isn't just asserted. perso-benchmark measures it directly: in-process WASM evaluation runs at ~1.75µs median, versus ~50.3µs for the same decision over a localhost network call to a PDP-style server — roughly 29× faster at the median, ~53× at p99. That's the best case for the network path (no TLS, no geographic distance, no multi-tenant queueing) — see the benchmark repo for full methodology, hardware specs, and raw results.
Python
Node.js
TypeScript
View more
Python
Node.js
TypeScript
Rust
Anthropic’s model context protocol
View more