Schedule a call

Case Study · MedTech · Irvine, CA

Lidavex

A Proof of Concept for a LiDAR-based medical measurement device: a desktop application that captures live scans from a real sensor, runs the client's proprietary algorithm on them, and makes every processing stage visible in real time — so the algorithm could be validated before a dollar went into device hardware.

  • Started May 2025
  • Team of 4
  • Fixed price · 0 deviation
  • Feature-complete in ~2.5 months
← Other industries

01 / The Challenge

Prove it before
you build it

Hardware is expensive to be wrong about. The mission of this PoC: answer the only question that mattered — does the measurement approach work on real sensor data — before Stage 2 (device prototype) and Stage 3 (manufacturing) were funded.

Algorithm
first

The client owns the algorithm IP and its clinical validity. Our job was faithful implementation and honest instrumentation — every processing stage rendered on screen, nothing hidden, nothing guessed.

Medical-device
discipline

An algorithm-validation stage in a medical device program: documented accuracy tests on physical reference objects, point-by-point responses to client findings, and full source-code rights transfer on completion.

Real hardware,
real time

A live sensor on a serial link, a UI that never blocks while data streams in, and reproducible offline replay of every captured scan — so any measurement can be re-examined later.

02 / Team & Process

Four people,
fixed price

4team members
3milestone payments · fixed price
≤3testing cycles in the contract
6documentation deliverables

Composition

  • 1 system architect / CTO
  • 1 system software engineer
  • 1 tester · 1 account manager

Cadence

  • Fixed-price engagement with milestone reviews
  • Requirements and delivery docs in Notion
  • Tasks and progress in Jira

Delivery

  • Feature-complete in ~2.5 months — within the contracted window
  • 0 budget deviation, no paid change requests
  • Out-of-scope findings triaged into a clean CR

03 / Technology

One app,
zero bloat

Python 3.12, a single lean desktop application — built to do exactly what the validation stage needed, and nothing it didn't.

  • ApplicationPyQt5 · threaded 4-plot layout with a control panel · Matplotlib rendering
  • ProcessingNumPy / SciPy signal pipeline behind the client's algorithm chain
  • DeviceVendor LiDAR SDK — native C++ extension built from source (CMake + SWIG + VS Build Tools) · USB-to-Serial link
  • DataTimestamped CSV log capture · interactive multi-file browser · offline replay of any session
  • Packagingpyinstaller → standalone Windows 10/11 executable, no installation required

04 / The Scale

By the numbers

30Jira issues
6epics
5Python modules
5processing algorithms implemented
8tunable parameters exposed in the UI
2operating modes: live capture · offline replay
5documented accuracy tests on reference objects
7mmedge measurement — within the 5–7 mm tolerance

05 / Outcome

Fixed price. Zero drift.

From kickoff in May 2025 to feature-complete at the end of July — within the contracted window, with no paid change requests. The algorithm was validated on real sensor data, accuracy findings were answered point-by-point, and the client received full source-code rights — ready to take Stage 2 forward.

"The entire team of SG Systems has been very quick, efficient and highly professional in every step of our work together. Our CTO has been impressed by the level of precision and speed."

Kamola Mir — CEO, Founder
Validate your idea with us → Other industries