Zipcode local coding agent
Zipcode helps teams understand, modify, test, and deploy code in environments where private source cannot be pasted into cloud-only assistants.
AI agent workflow platform
Petaflo builds secure AI workflows for teams that cannot send everything to the cloud.
Our first product, Zipcode, is a local and edge AI coding agent for privacy-sensitive, low-connectivity, and air-gapped engineering environments.
Founder-led
Founded and operated by Byungwon So in Seoul, South Korea.
Secure AI product
Built for private code, controlled sync, and auditable agent runs.
B2B focus
Regulated engineering, infrastructure, software, and operations teams.
Built for AI work that needs local context, audit trails, and controlled cloud execution.
Petaflo turns sensitive engineering and operations workflows into tracked AI agent runs with source evidence, approval points, and exportable deliverables.
Outputs with receipts
Every recommendation keeps source references, assumptions, confidence notes, and open questions so reviewers can trust what changed and why.
Built to scale responsibly
Queue-based jobs, retrieval, provider routing, audit logs, and cost monitoring are planned from the start so AI usage can grow without becoming opaque.
Public product signals for partner review.
Petaflo is preparing pilot workflows around private-code assistance, evidence-backed reports, and secure deployment support.
Local code review packet
Zipcode scans a repository locally, summarizes architecture, identifies deployment gaps, and produces a review packet without exposing raw source code to public services.
Traceable run history
Each run stores prompts, model choices, source references, generated changes, test notes, and approval state for later audit and customer delivery.
Founder-operated startup
Petaflo is a Seoul-based, founder-led software startup founded in 2026 and reachable at the company-domain address admin@petaflo.com.
A practical AWS architecture plan for secure AI agent workflows.
Petaflo will use cloud infrastructure for real product workloads: hosted APIs, model evaluation, encrypted artifact storage, source retrieval, job execution, observability, and secure customer intake.
The AWS build plan includes Amazon Bedrock for model evaluation, EC2 and ECS/EKS for agent workloads, S3 for encrypted artifacts, ECR for container distribution, IAM/KMS/Secrets Manager for security, and CloudWatch for observability.
AWS
Bedrock, EC2, ECS/EKS, S3, ECR, IAM, KMS, Secrets Manager, and CloudWatch.
MVP
Repository intake, run history, source evidence, reports, and exports.
90d
Initial build window for Zipcode pilots, model tests, and secure cloud sync.
B2B
Teams that need AI assistance without losing control of private source data.
Roadmap with visible milestones.
The site, domain email, privacy notice, terms, and demo signals are the public foundation. The next steps move Petaflo into pilot-ready workflow software.
Public company surface
Live website, company-domain email, privacy notice, terms, AWS architecture plan, and public demo workflow signals.
Hosted workflow MVP
Build repository intake, source uploads, run records, status views, and report export so pilots can run through a consistent product flow.
AI execution layer
Add retrieval, evaluation traces, queue workers, provider routing, AWS cost dashboards, and secure artifact sync.
Petaflo is open for pilots and cloud partner review.
For startup program verification, pilot discussions, or workflow scoping, use the company-domain email below. Founder: Byungwon So. Location: Seoul, South Korea.
Company contact
admin@petaflo.com