traian@home : ~ / about×
guest@traian:~$
Traian Popescu

Traian Popescu

Software & founding engineer

I build software for a living — currently a product engineer at Soilytix, and a founding engineer before that. I like building things from 0 to 1. I started early: competitive C++ (USACO) and freelancing for Romanian clinics & brands while still in high school. Born in Bucharest, built in Amsterdam, strengthened in the USA.

On the side, I build my own projects, shoot photos, and tinker with split ergonomic keyboards. I once designed my own language too — miniC, a small compiler down to x86-64 — and I've traveled across five continents.

Age 22 → 23 on Jul 01Edu BSc CS, VU Amsterdam · minor at UoU
FlutterTypeScriptReact / Next.jsDartPythonLLMs
~/work — experience & education
experience

Product engineer at a soil-intelligence company. We turn soil data — microbiome sequencing, lab measurements, and geospatial layers — into decision tools for crop-science and agri-input R&D teams. I work across the whole stack, from the database to the map on screen.

  • Build data-heavy dashboards — interactive maps, time-series, and reports.
  • Help move the product from a project-centric monolith toward a modular platform.
  • Own features end to end: spec → ship → iterate, keeping the architecture clean as we scale.
  • TypeScript, React & tRPC on the front; Python services and spatial data behind it.
  • Visit soilytix.com ↗
TypeScriptReacttRPCMapLibrePython

Founding engineer at Vulture Labs, a San Francisco startup. The product was an AI layer on top of the CCTV a business already runs — no new cameras. We streamed every feed into a small on-premise NVIDIA box and ran our own tech right there, so the footage never left the customer's site.

  • Turned ordinary camera frames into 3D scenes — a metric digital twin of the site — with our own brain-inspired monocular-depth model (ThirdEye).
  • Made all that footage searchable: every frame embedded with a vision-language model (SigLIP) into a vector database, so you could find a moment by describing it or attaching an example image.
  • Hit a wall with plain similarity search across cameras, so we pivoted to a knowledge graph on an intelligence-style ontology — people, vehicles, events and how they connect — that could reconstruct what happened and in what order.
  • Put an LLM on top (through MCP, brand-new at the time) so you could just ask — “what delayed this morning's delivery?” — and get an evidence-backed timeline back.
  • Saw it used for real: delivery & logistics forensics for enterprises, theft-suspect alerting for shops.
  • Then we went after the other half of privacy. Running on a box on-site keeps footage local, but you can't put a box in every home, and the cloud normally means handing your video to someone else. So we found a way to run the whole pipeline in the cloud with everything encrypted. Each customer's data stays mathematically isolated, even when it sits on the same storage as everyone else's. We filed a patent on it. That's what made the consumer product possible.
  • That unlocked consumer: I owned the move, building the mobile app from scratch (Flutter, clean architecture, BLoC) to bring the same intelligence to smart-home cameras like Ring.
  • Visit vulturelabs.com ↗
  • A sneak peek at what we were building ↗
Computer VisionEdge AILLMs / MCPKnowledge GraphsFlutter

Built the Engie mobile app — a consumer app for one of the largest electrical utility companies, with 700+ monthly active users.

  • Shipped features for an app with 700+ MAUs at one of the largest electrical utilities.
  • Architected scalable, maintainable code with Clean Architecture & MVVM.
  • Worked closely with Functional Analysts, Product Designers & QA to nail specs, usability and quality.
  • Embraced a Shift-Left approach to testing and prioritized accessibility from day one.
  • Reached 100% test coverage across features — unit, widget & integration tests.
  • Visit inthepocket.com ↗
FlutterDartClean ArchMVVMTesting

A full year, full-time alongside my CS degree — building a centimeter-accuracy GPS surveying app for high-precision positioning hardware.

  • Built a cross-platform (iOS/Android) surveying app in Dart, Flutter & Firebase, Agile throughout.
  • Integrated REST APIs in a microservice ecosystem and a WebSocket channel for real-time RTK GPS data.
  • Migrated to a feature-first, domain-driven architecture with BLoC — more modular, more scalable.
  • Added Firebase Analytics & Crashlytics; designed unit, behavioural & integration tests.
  • Collaborated with designers, PMs and the embedded/cloud teams to stay aligned.
  • Held a real engineering job full-time while doing a full CS degree.
  • Went deep on Clean Architecture, BLoC & domain-driven design — all by hand, back when ChatGPT wasn't really a full thing yet. :)
  • Visit marxact.com ↗
FlutterDartFirebaseBLoCWebSocketsGPS / RTK

TA for Data Structures & Algorithms on the AI track — from sorting to advanced graph algorithms.

  • Led instruction for 400+ AI students, from sorting through advanced graph algorithms.
  • Guided students using Python for algorithmic development and implementation.
  • Reviewed code, gave feedback, and supported debugging & optimization.
PythonAlgorithmsTeaching

Where it started — in high school, building websites and apps for Romanian private clinics and brands. My first taste of shipping for real clients.

  • Rafiziomed — presentation site for a private clinic, built with React, ES6+ JS, HTML & CSS.
  • Where I learned state management & the React fundamentals the hard way — components, props, hooks, routing & forms.
  • Figured out the whole deployment path end to end: builds, domains, DNS & hosting, all the way to live.
  • Delivered sites & apps for paying clients as a teenager — and learned to scope, ship & talk to non-technical people.
  • All self-taught the old way — docs, forums & trial-and-error, long before AI could just hand you the answer.
ReactFlutterWebFreelance
education
Vrije Universiteit Amsterdam · BSc Computer Science
Selected coursework: Data Structures & Algorithms, Computer Organization, Operating Systems, Computer Networks, Automata & Complexity, Machine Learning, Software Design, Software Engineering Processes, Human-Computer Interaction, Databases, Data Wrangling, Linear Algebra.
2026
University of Utah · Minor · semester abroad
Coursework: Algorithms, Parallel Programming, Mobile App Programming, Image Processing Basics & Ethics in Data Science.
2024
achievements
USACO · competitive programming
C++ algorithmic contests in high school — 1000/1000 at Bronze, 683/1000 at Silver.
HS
~/making — projects & papers
projects
papers
building…
~/blog — compiling posts
first-post.mdx
building…
~/life — rendering memories
photos/ travel/
~/contact — say hi
email

Open to conversations.

guest@traian:~$
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
-- NORMAL --"" [EN] 1,1 All
● onlineuptime: 22 yearsnext build: v23.0 — Jul 01~/about