Skip to content
Melih Onat

Product engineer

Applied AI · Full stack

Toronto,  

Melih Onat, product engineer. I build software people rely on: the retrieval behind public AI assistants in around twenty German municipalities, a legal workflow used every day, and Resonate, a music platform I created and run on my own.

Open to full-time roles in product engineering and applied AI. Based in Toronto and authorized to work in Canada.

Selected work

Two systems in production and one experiment. Every figure on this page comes from something that shipped.

01In production2025 — present

Resonate

A social music platform built around listening to albums together.

Resonate album page for In Rainbows by Radiohead, with the tracklist and rating controls
An album page: catalogue data, ratings and reviews in one place.
registered users
~200
ratings submitted
4,000+
API routes
25+
paid acquisition
0

Streaming made nearly every album available and listening increasingly solitary. Resonate’s Listening Club sets a shared pick and a common listening window; members listen, rate and talk about the same record, and the catalogue keeps what happened. I own all of it — product direction, interface, database, real-time messaging, deployment and user support.

Stack: Next.jsTypeScriptSupabasePostgreSQLVercel

02In production2024 — 2025

Neuraflow

Retrieval infrastructure for public AI assistants in around twenty German municipalities.

City of Bonn website with its municipal AI assistant open, answering an identity-card question
The assistant on the City of Bonn’s website, answering an identity-card question.
municipal customers
~20
data refreshes
Weekly
retrieval per municipality
Isolated
production deployments
Public

As the first developer hire, I worked with the CTO on the shared platform behind public-facing municipal assistants: ingesting XML APIs, websites and documents, splitting them into focused retrieval units, keeping each municipality’s data isolated, and shipping the retrieval, prompting, evaluation and infrastructure around it.

Stack: PythonOpenAIPineconeAzureDockerTerraform

03Lab note2024

whatsyourtune

Computer vision translated into music.

An experiment that answers a face with piano. A happy expression gets quick major-key phrases, a sad one slower minor-key lines. TensorFlow reads the expression, Magenta generates the music, and the loop stays fast enough to feel like a single interaction.

Stack: PythonTensorFlowMagenta.js

whatsyourtune detecting a happy expression and generating music in response
The model reads an expression; the music engine answers.

Experience

  1. 2026 — present

    Product Engineer

    Salubrum, Toronto

    Building an audience-building platform for US healthcare clinics: developed its demographic filter and reusable React and TypeScript components, turning UX requirements into product features across front end and back end.

  2. 2025 — present

    Creator & Product Engineer

    Resonate, Toronto

    Designed, built and run a social music platform end to end: product direction, interface, API and data design, real-time messaging, deployment and user support. Case study

  3. 2025

    AI Developer Intern

    Humber College, Toronto

    Built an AI hint assistant for an escape-room installation, connecting NLP pipelines, MongoDB game state and real-time player guidance.

  4. 2024 — 2025

    Generative AI Engineer

    Neuraflow, Bremerhaven

    First developer hire. Worked with the CTO on the shared RAG platform behind municipal assistants, from ingestion and retrieval to evaluation and deployment. Case study

Education

2026
Master of Digital Media, Toronto Metropolitan University
2025
B.Sc. Digital Media (Media Informatics), University of Bremen

Toolkit

Languages
Python · TypeScript · JavaScript · SQL
Product
React · Next.js · Node.js · Flask · REST APIs
AI & data
RAG · LLM integration · Embeddings · Pinecone · PostgreSQL · Supabase · PyTorch · MCP
Infrastructure
Docker · Azure · GCP · Vercel · Fly.io · Terraform · Redis · CI/CD

About

Melih Onat outdoors, by a river
Melih Onat. Toronto, 43.65° N.

There’s a person behind the stack.

I care about what gets built, not just whether it works. Good software is well made, useful, and pleasant enough that people want to keep using it.

I studied media informatics at the University of Bremen and finished a Master of Digital Media at Toronto Metropolitan University in 2026. I work in English and German.

Coding agents are part of how I work every day — for implementation, debugging and refactoring. The problem definition, the architecture, the review and the production result stay with me.

Off the clock

GeoGuessr
Road markings, utility poles, languages, and the occasional lucky guess.
Music
Albums, recommendations, and the conversations that became Resonate.
Classic books
A slower activity to balance out a day spent coding.
Gym
The simplest reset after too many hours at a desk.

Contact

Toronto,  
 

Tell me what you’re trying to ship.

hello@melihonat.dev

Open to full-time roles in product engineering and applied AI. Based in Toronto and authorized to work in Canada.