William Lo Channiko

Applied AI + Full-Stack Engineer

William Lo Channiko

I build AI systems for evidence-heavy work, with software-enforced rules and explicit human decisions.

View projectsDownload resume
William Lo Channiko

AI should support the decision, not hide it.

Models contribute inside a workflow whose context, rules, and decision authority stay visible.

  1. ContextSource material stays visible
  2. AI proposalBounded, reviewable contribution
  3. Software checksDeterministic rules own constraints
  4. Human decisionAuthority remains explicit

AI Applications

Document and case-resolution workflows, from source evidence and local inference to human-approved outcomes.

Invoice Review

Human-approved invoice review and ERPNext drafts

Invoice Review workspace showing a synthetic invoice beside extracted fields, validation findings, and an approval blocker.
Review workspace with synthetic fixture data

ERPNext workflow benchmark complete

An AI invoice-to-ERP workflow that extracts PDF data, checks it against deterministic rules, and gives a reviewer the evidence and authority to approve a controlled ERPNext draft.

68%
Lower median elapsed invoice-to-ERP draft time (153s to 49s)
10/10
Expected assisted-workflow results, versus 9/10 through direct entry
98.75%
Exact field match on a sealed 10-invoice licensed synthetic holdout (79/80)

Current boundary

One operator used synthetic invoices in a provisioned local ERPNext sandbox. The 68% compares elapsed time, not separately measured active work. Azure integration validation used mock providers and was temporary; newer reliability scenarios remain local, not Azure-live.

View case study
Repository

Case Resolution Copilot

Policy-governed decision workspace

Hosted Case Resolution Copilot Decision Brief for synthetic case CS-DEMO-001, showing verified facts, a suggested resolution, uncertainty, and human-review guidance.
Decision Brief captured from the hosted demo with synthetic case data on October 12, 2026

Controlled-pilot readiness gate passed

A policy-governed AI case-resolution system that converts fragmented evidence and versioned policies into review-ready Decision Briefs, with deterministic risk controls and human approval for consequential actions.

83.7%
Lower raw median case-resolution workflow time (582s to 95s)
3/3
Complete safe workflows, versus 0/3 through the manual path
20/20
Synthetic regression observations passed

Current boundary

The benchmark used three matched synthetic cases and one developer operator. The October 12 hosted screenshots show a read-only product tour, not workflow acceptance. Controlled-action and Gmail-draft acceptance were recorded separately. AWS validation was temporary; the always-on demo remains on Vercel and Neon.

View case study
Repository

Private Offline Document Intelligence

Local, cited document answers and approved Odoo checklists

Synthetic UI replay of a cited answer beside its source document page.
Synthetic UI replay of the validated workspace, not a live-model capture

Offline local lab handoff validated

A local workspace helps MSP teams search client contracts and SOPs, answer with page-level sources, and prepare service checklists for separate review before saving them to Odoo.

13/13
Synthetic answer and abstain/review regression cases passed
5.056s
First sourced answer on fresh local volumes
0
Published service ports in the rechecked Compose profile

Current boundary

Services used an internal Docker network and temporary loopback bridge; the host was not physically air-gapped. The results do not establish customer accuracy, long-running uptime, or IT change execution. The updated no-port source profile needs a newly validated release bundle.

View case study
Repository

Professional experience

Full-stack engineering inside an operating business.

Two full-time internship terms covering native Android, offline workflows, APIs, an internal web system, approval flows, imports, and business reporting.

PT Dover Chemical

Full-Stack Software Engineer Intern

Built internal sales-operations software for PT Dover Chemical across the CRM Dover Chemical Android app and a Customer Management System, covering offline synchronization, approval workflows, master-data management, imports, and analytics.

Jan-Aug 2025

CRM Dover Chemical - Android App

  • Owned end-to-end development of an offline-first Draft Sales Order application, from requirements gathering with IT supervisors and sales users through implementation, integration, and testing; built local persistence, background synchronization, caching, retries, and attachment handling across 38 API operations and 15 Room entities/DAOs.
  • Kotlin
  • Room
  • Retrofit
  • WorkManager

Aug 2025-Jan 2026

Customer Management System - Web

  • Independently designed and built a Customer Management System end to end around a 6,656-record customer master dataset, translating sales and administrative requirements into validated imports, reporting, audit trails, and role-based approval, rejection, and reopen workflows across 15 data domains.
  • PHP
  • CodeIgniter
  • MariaDB
  • Bootstrap
  • jQuery
  • Chart.js
  • PHPExcel

Engineering AI as a system, not a feature.

I am an early-career engineer based in Indonesia with full-stack experience across Android, web, APIs, and business data. I build AI workflows and validate AWS/Azure deployments with a focus on reliability, monitoring, and operational readiness. My end-to-end ownership covers architecture, implementation, evaluation, deployment validation, and recovery.

Experience
1 year as a full-stack software engineer intern
Based in
Indonesia (UTC+7)
Education
Bachelor's Degree in Informatics, graduated with Distinction
Current focus
Document operations and support decision systems

Let’s build software people can trust.

Open to Applied AI and full-stack engineering opportunities with teams that value product clarity, measurable verification, and honest system boundaries.

Email WilliamDownload resumeLinkedInGitHub