Hardline Engineering case study Source

Source-safe portfolio case study

From phone capture to a trustworthy wall record.

Hardline is an engineering system for documenting exposed utilities before walls close. This public artifact shows the product thinking, system boundaries, and documented work of co-founders Keenan Francisco and Caleb Ponce without publishing the production codebase.

Visit the Hardline product site Explore the synthetic demo

Evidence boundary: the visualization uses a code-rendered scene and synthetic utility data. It is not a field-accuracy result, customer scan, or performance claim.

SYN-042 / OPEN FRAMING / NORTH WALL
SYNTHETIC RECORD
Synthetic mapped utilities A reference visualization containing studs, two wire paths, one pipe, and three outlet markers.
Interactive product-language sample. All detections and coordinates are synthetic.

Private production systemiOS, Python, TypeScript

514 founder-authored commitsGitHub API, 2026-09-08

Public artifactsynthetic data only

The system

One field workflow across three engineering surfaces.

The production system connects an iPhone capture client, an asynchronous processing backend, and a browser review console. This diagram deliberately stops at architectural boundaries; implementation and infrastructure remain private.

  1. 01
    Capture

    Phone-guided recording with spatial context and field feedback.

    Swift / ARKit
  2. 02
    Process

    Queued reconstruction, detection, provenance, and record assembly.

    Python / FastAPI
  3. 03
    Review

    Permissioned 2D and 3D inspection, correction, export, and sharing.

    TypeScript / React
  4. 04
    Return

    Alignment-aware guidance back to the documented physical wall.

    Swift / spatial math

Engineering decisions

The hard work was deciding what the system is allowed to claim.

The production history records a shift from feature breadth toward honest geometry, observable failure modes, and a review trail that distinguishes measurement from inference.

Selected authored work / Keenan

Keep measurement and inference separate.

Coordinate provenance travels with the result. Measured wall-local positions, approximate image-relative locations, and unreferenced detections cannot silently collapse into one confidence claim. 

position_basis
  measured        → wall-local frame
  approximate     → image-relative only
  unreferenced    → no position claim

Selected authored work / Caleb

Make the record usable beyond the scan.

The workflow became a full product surface: durable room identity, guarded uploads, workspace access, 2D review, controlled sharing, quotas, and field-facing iOS feedback. 

record_lifecycle
  capture → upload → workspace
  review  → share  → return

  boundaries: auth / rate / quota

Shared product discipline

Treat a plausible picture as the beginning of validation, not the end.

Synthetic tests verify software behavior. They are not field evidence. Hardline's validation contract requires physical ground truth before any measured accuracy claim.

Synthetic replay software check Measured pilot field observation Reviewed report claim boundary

Contribution record

Built by Keenan Francisco and Caleb Ponce.

Counts below are attributed commits reported by GitHub for the private production repository on September 8, 2026. Selected work was internally verified against authored commit evidence; outside reviewers cannot inspect those private objects, and attribution does not imply sole ownership of shared product decisions.

Keenan Francisco

@keenanfrancisco

286authored commits

Reliability, spatial contracts, field validation, iOS capture, and deployment safety.

  • Reworked capture-to-return coordinate and datum contracts across backend, web, and iOS.
  • Built adversarial evaluator baselines and field-evidence reporting paths.
  • Added wall-coverage gating, pipeline observability, secret scanning, and release CI.
View GitHub profile

Caleb Ponce

@calebponce

228authored commits

Full-stack product delivery, reconstruction integration, security, sharing, and field UX.

  • Integrated the RGB-only reconstruction path and connected it to scan processing.
  • Shipped durable room identity, public review links, and the 2D utility record.
  • Hardened authentication and rate boundaries while closing iOS scan-flow gaps.
View GitHub profile

The internally verified contribution methodology and selected private commit references are documented in CONTRIBUTIONS.md.

Disclosure boundary

Public enough to inspect. Narrow enough to protect the company.

Included here

  • Interactive synthetic product-language demo
  • High-level system boundaries and technology choices
  • Dated GitHub contribution counts for both founders
  • Selected engineering decisions and evidence boundaries
  • Runnable, dependency-free portfolio source

Kept private

  • Production application and complete Git history
  • Core reconstruction, geometry, and relocalization implementation
  • Infrastructure configuration, secrets, and operational history
  • Customer, property, scan, or field-validation data
  • Internal go-to-market and company planning documents

Product site / private walkthrough

Ask either founder for a private architecture walkthrough.

Visit the product site Keenan Caleb