Skip to content

Vol. 1September 26, 2026Charlotte Edition

Vedi Gharibian

Builds software. Runs live operations.

Software Engineer · Charlotte, NC

Scroll to continue

Toolkit

Fig. 1.1: Vedi Gharibian

About

A software engineer in Charlotte building full-stack platforms and real-time systems.

Vedi Gharibian is a software engineer in Charlotte, North Carolina, building full-stack systems in TypeScript, Node.js, and Flutter on a live e-commerce platform, and real-time computer vision in Python. A 2023 computer science graduate of UNC Charlotte, Gharibian has moved into LLM engineering, working across production platforms, API security, and applied computer vision.

The work spans two sides of the same discipline: building systems and probing them. On the product side, that means the Flutter application sellers use to run their storefronts, the marketplace services behind it (vendor onboarding, catalog and search, multi-vendor checkout, delivery), and the migrations, role-based access control, and Arabic/English localization that come with changing a platform already serving users. On the security side, it means API security assessments, including a GraphQL review of a mobile application backend.

Alongside the engineering track, Gharibian runs live event operations for a county venue hosting about 60 events and more than 120 event-days a year. The schedule is published, the timing and scoreboard systems have to work on the first attempt, and the building is full while it happens. It is the same standard as shipping to production, on the same calendar.

Building a system and taking one apart are the same skill pointed in opposite directions.
Gharibian

Section B

Background

The things I work with.

Software

  • TypeScript
  • Node.js
  • Python
  • .NET
  • Fastify
  • MongoDB
  • Angular
  • Flutter
  • AWS S3

AI and computer vision

  • Python
  • FastAPI
  • YOLO
  • Face embeddings

Security

  • API security testing
  • GraphQL

Event technology

  • Colorado Timing System
  • Swiss Timing
  • Daktronics OmniSport
  • Hy-Tek Meet Manager

  • 12

    Cyrys services, down from 20

  • 60

    events a year

  • 120+

    event-days a year

Progress Report

In more detail.

  1. 01December 2023

    Foundations

    B.S. in Computer Science from UNC Charlotte, with a Software Engineering concentration and a Mathematics minor. The degree is the base the rest of the track is built on.

  2. 02Present

    Production software

    Software engineer at Elryan, a multi-vendor e-commerce marketplace serving Iraq, held concurrently with the county role. Full-stack TypeScript, Node.js, MongoDB, and Flutter on a live platform. The mobile side is the vendor application, the Flutter client sellers use to manage listings and inventory and work through incoming orders; the platform side runs from vendor onboarding and management through the product catalog and search, cart and checkout across multiple vendors, and delivery and logistics. Underneath all of it: role-based access control, Arabic/English localization, and a run of migrations: framework and version upgrades, a restructuring into clearer modules, and schema changes against live data.

  3. 03

    Security

    API security testing, including a GraphQL security assessment of a mobile application backend. The work moved the engineering track from building systems to probing them.

  4. 04

    Computer vision

    Built Cyrys, a real-time face detection and tracking system, out of a long-standing interest in the field.

    See the Cyrys report

  5. 05Now

    LLM engineering

    Current

    Building AI-powered product features and tooling for automated codebase audits, having moved into LLM engineering on a base of full-stack JavaScript and some .NET.

/resume

Section C

Selected Work

Special Report
  • Python
  • FastAPI
  • YOLO
  • WebSockets
  • Vue

Cyrys

Face detection, person tracking, identification, and automated watchlist monitoring, built as Python microservices.

The Cyrys pipelineA pipeline diagram. A video stream feeds a detector. The detector hands work to asynchronous face processing, which passes results to a unified deduplication service, then to a watchlist check, which emits alerts over WebSockets. Analytics branches off the detector as a separate path.Video streamDetectorAsyncface processingUnifieddeduplicationWatchlist checkWebSocket alertsAnalytics
Fig. 2.1: The Cyrys pipeline

Cyrys reads a video stream, detects and tracks people in it, identifies faces against stored embeddings, and checks what it finds against a watchlist. The system is in active development, and most of the work so far has been architectural: making a pipeline that has to keep up with a live stream actually keep up with it.

The service count came down from 20 to 12. The reduction came from removing redundancy and improving parallel processing, and it had a second effect that was not obvious up front: fewer services writing to disk at the same time meant less lock contention between them.

Three separate deduplication systems (face-based, automatic, and profile-enhanced matching) were merged into a single service behind one API, without giving up profile-based matching accuracy. Watchlist monitoring moved into the stream-processing pipeline itself, with automatic face checks, live alerts over WebSockets, and cooldowns so a single subject cannot flood the alert channel.

The largest real-time win was moving face processing off the main detection thread and onto asynchronous processing. Throughout the refactor, the public API stayed compatible, with backward-compatible endpoints and migration guides.

  • 20 → 12 services
  • 3 → 1 deduplication systems
  • Live WebSocket alerts
  • Async face pipeline

Day Edition

The Operations Desk

Live event operations

2+ years

A second, concurrent track: running live, high-stakes event operations. The schedule is published, the timing and scoreboard equipment has to work on the first attempt, and the building is full while it happens. It is the same discipline as shipping to production, and it runs on the same calendar as the software work.

  1. Runs a venue hosting about 60 events and 120+ event-days a year, including competitive swim meets.

  2. Operates swim meet timing and scoreboard systems: Colorado Timing System, Swiss Timing, Daktronics OmniSport, and Hy-Tek Meet Manager.

Section D

Editor’s Note

The work Gharibian points to first is the Cyrys consolidation: twenty services reduced to twelve, and three separate deduplication systems merged into one, without giving up the profile-based matching accuracy that made the system worth building. Cutting complexity is easy when you are willing to lose something; the constraint was losing nothing.

Section E

Let’s Talk

I read everything that comes in, and I answer anything specific: a role, a system you are building, or a problem you think is mine to look at.

Vedi Gharibian

Software Engineer

Email me
…

registered agent accesses