Oluş Emre Demir

Product, UX, DFM and embedded systems portfolio of Oluş Emre Demir.

NAME
Oluş Emre Demir
LANGUAGE
ENGLISH / EN
WEBSITE
OLUSEMRE.DEV
CONTACT
OLUSEMREDEMIR@GMAIL.COM

[01] Projects & evidence

Gage / 2026

I built Gage to inspect why a part image received an anomaly decision. I put the score beside a normal reference and a difference map, then added a way to record a reviewer’s correction.

Status
exploration
Personal contribution
I built the reference comparisons, experiment and local review interface.
Evidence / limit
I compared two CPU methods on generated plate images and retained the failures under lighting and position changes.

Mimo / 2026

I built Mimo to answer a preference question after the preference changes, without losing when the earlier statement applied. I stored original messages locally and made person, conversation, scope and valid time part of the answer.

Status
exploration
Personal contribution
I built the local memory model, temporal resolution, three-method comparison and query interface.
Evidence / limit
All three methods tied on the repeated coffee-preference fixture; a separate correction diagnostic exposed a stale lexical answer.

Flick / 2026

I built Flick to distinguish gestures by their paths through time, even when they pass through a similar hand pose. I process circle, swipe-right and zigzag trajectories and connect an accepted gesture to a visible in-app action.

Status
exploration
Personal contribution
I built trajectory preprocessing, two classifiers, the trigger state machine and local replay.
Evidence / limit
I compared DTW and RandomForest on 48 generated test clips, inspected a rejected zigzag and exercised one-action behavior for a held gesture.

Cue / 2026

I built Cue to take a search result beyond a video name into an inspectable time interval. Each result shows a frame, transcript and the ranking contributions from image and text channels.

Status
exploration
Personal contribution
I built windowing, two-channel retrieval, rank fusion, the CLIP comparison and the local results interface.
Evidence / limit
I compared methods on six answerable fixture queries and retained the model’s false evidence for an absent scene.

Human Industries / 2026

I designed an identity and website for Human Industries around my interest in mechanisms. I wanted a place to watch movement, inspect connections and try different arrangements.

Status
design-development
Personal contribution
I designed the identity, website and flywheel concepts.
Evidence / limit
I built the website and mechanism lab. I am continuing to develop the flywheel concepts.

Talveniq / 2026

I designed a new identity and interactive website for Talveniq. I turned the idea of voice and text agents into example conversations visitors can inspect step by step.

Status
exploration
Personal contribution
I designed the identity and interface, and built the website.
Evidence / limit
I published the website and built the product flows as browser-based simulations with sample data.

BL01 BalanceLab / 2026

A desktop ball-balancing instrument designed to make the relationship between mechanics, sensing and control visible. Three coordinated motor stations move a circular plate above a clipped triangular enclosure.

Status
coursework
Personal contribution
Mechanism and product design documented in the project report.
Evidence / limit
Course report and design artifacts; simulation and physical-system results require separate evidence.

Demir Works / 2026

Demir Works is a brand and web platform that turns management consulting for manufacturing and operations teams from an abstract promise into a visible operating system. I built the product around finding the real constraint, connecting each decision to one owner, and closing an action only with evidence from the field—not around asking people to work more.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

Sealed Stories / 2026

Sealed Stories is a hybrid fantasy narrative platform that combines physical novel and tabletop role-playing experience with a living digital universe called Animaeris. While optional QR codes in the book open to mini-story, drawing, sound and game layers, the web experience takes the story beyond the physical page with world atlas, lore archive, character generation, Storyteller sessions and community layers.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

Flâneuse Mag / 2026

Flâneuse Mag is a bilingual art platform that brings together contemporary artists, interviews, curatorial articles, digital magazine issues and browser-browsable online exhibitions in a single editorial universe.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

LinkagePy / 2026

LinkagePy is a Python-based kinematics working environment that I developed to build two-dimensional mechanisms, solve their movements, and examine their behavior in real time. The project is not just a demo showing the inverse kinematics result; It combines mechanism modeling, forward and inverse solution, degree of freedom analysis, motion generation, waypoint planning and engineering graphics in a single desktop interface.

Status
working-prototype
Personal contribution
Mechanism model, numerical solver, desktop interface and analysis tools.
Evidence / limit
Public source and interface screenshots are linked. Historical test results are not a current release certification.

Lurneo / 2026

Lurneo is a web-based learning application that turns user-selected supported YouTube videos into English learning materials that include synchronous bilingual subtitles, in-context vocabulary study, speaking/shadowing repetitions, personal learning package and quiz.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

WhiteGate / 2026

WhiteGate is an active AI transformation studio where we transform companies' dispersed operations into systems that work with custom software, artificial intelligence agents, automations, integrations and decision panels.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

2yaka / 2026

2yaka is a multi-faceted mobility platform that transforms the paper book, WhatsApp list and verbal agreement at taxi stands into a live, recorded and role-based operation system.

Status
released
Personal contribution
Product architecture, mobile apps, operations panels and backend integration.
Evidence / limit
Build and deployed interfaces are documented. Field impact, revenue and dispatch reliability need a dated dataset.

602 — Earthquake Aid Kit /

602 is a compact earthquake aid station named in direct reference to 6 February. It began with a personal question after the earthquakes: could essential communication, light and power be brought into one product that people could reach easily and trust when normal infrastructure failed?

Status
design-development
Personal contribution
Product concept, mechanical packaging and proposed embedded interaction.
Evidence / limit
Design and packaging images; a functioning integrated aid device is not demonstrated.

Airium / 2021

Airium was my first complete product attempt: a desktop air purifier that moved from CAD to 3D printing, electronics, assembly and sensor-backed testing.

Status
design-development
Personal contribution
Mechanical product design and internal module packaging.
Evidence / limit
CAD and enclosure views; manufacturing and device performance are not demonstrated.

Purge One / 2024

Purge One started as a course assignment about building an experimental environment and its instrumentation, then became a compact filament dryer with a controlled heater, ESP32 logic, airflow routing and lab calibration.

Status
coursework
Personal contribution
Filament-dryer concept, enclosure, airflow guide and instrumentation design.
Evidence / limit
Coursework and CAD are shown. Original measurement records and efficiency comparison are not attached.

talks2me / 2026

talks2me is a live web experience that turns an everyday object on camera into a Turkish-speaking character for 90 seconds.

Status
Not stated
Personal contribution
See case narrative
Evidence / limit
Artifacts in the linked case; independent outcome data is not stated.

Scrill / 2026

Scrill is a small menu bar application on macOS that adds sound and Force Touch haptic feedback to each scroll detent without changing the scroll movement.

Status
working-prototype
Personal contribution
Native macOS utility, scroll-event handling, sound/haptics and release packaging.
Evidence / limit
Interface and build work are documented; notarization, purchase and physical Mac acceptance are not evidenced here.

[02] Labs

Gage

I built a normal-reference bank and compared new part images through heatmaps.

AI / ML study / Exploration

I ran the experiments on generated fixtures and describe the decisions and results in the case study.

Mimo

I built local preference memory with person, conversation and valid-time boundaries.

AI / ML study / Exploration

I ran the experiments on generated fixtures and describe the decisions and results in the case study.

Flick

I classified hand trajectories and turned accepted gestures into debounced in-app actions.

AI / ML study / Exploration

I ran the experiments on generated fixtures and describe the decisions and results in the case study.

Cue

I built a search tool for finding moments from frame sequences and timed transcripts.

AI / ML study / Exploration

I ran the experiments on generated fixtures and describe the decisions and results in the case study.

Acoustica

A Python tool for examining room acoustics.

Python tool / Exploration

Source code is available to inspect. Acoustic accuracy and calibration results are not established here.

Fall model

A numerical physics model in PhysicsSim’s fall_sim.py.

Numerical model / Exploration

The code is linked for model inspection; a comparison with experimental measurements is not published here.

Caster digital twin research

Research into modeling drive wheels and a freely swiveling caster.

Mechanical research / Research only

This is a research repository. No executed simulation, calibrated contact model or physical result is claimed.

Numerical mechanism solver

Separating geometry and numerical solving from LinkagePy’s interface loop.

Method note / Documented method

This note points to the existing LinkagePy case. Historical test counts are not presented as a fresh run.

Scroll detent engine

Turning trackpad and mouse scrolling into audio and haptic triggers.

Native interaction / Documented method

The method is described in the Scrill case. This illustration is not evidence of a native app or haptic test.

Airflow guide

Treating Purge One’s filament cradle as part of the air channel.

Thermal design method / Coursework

The geometry comes from the existing CAD and case narrative. No efficiency improvement is claimed without sourced measurements.

[03] Practice

About

I am Oluş Emre Demir, a product and UX engineer working across mechanical constraints, embedded logic, interface systems and early product strategy.

I like projects where the answer cannot live in only one layer. A useful product has to make sense as an object, as a system, as an interface and as a thing someone can actually build, ship, maintain and use.

My work usually starts with a rough idea, a technical constraint or a messy workflow. I turn that into a clearer product model, then move toward prototypes, interface flows, implementation details and the decisions that make the product easier to manufacture, explain and operate.

What I build

I build product systems, UX flows, internal tools, technical interfaces and prototypes for hardware-adjacent products. I am comfortable moving between a product map, a screen, a firmware behavior, a manufacturing constraint and the small details that make the final experience feel coherent.

I care about products that survive contact with reality: interfaces that match the operator workflow, components that respect DFM decisions, dashboards that expose the right state, and prototypes that answer the riskiest question instead of only looking finished.

How I work

I work from first principles. I clarify the user, the job, the physical or technical limit, and the failure mode before I polish the surface. Then I reduce moving parts, prototype the uncertain layer and make the product language simple enough that engineering, design and business decisions can point in the same direction.

I prefer small loops with real evidence: sketch the model, build the minimum working surface, test the interaction, remove what does not matter, and keep tightening the relationship between product logic and implementation.

Where I am useful

I am useful when a product sits between disciplines: a device needs a usable control surface, a technical workflow needs a clearer interface, an early venture needs to find the shortest path from concept to market signal, or a team needs to translate manufacturing constraints into product decisions.

My range is strongest around product thinking, UX systems, prototyping, DFM strategy, embedded behavior, interface design and hands-on delivery. I do not treat those as separate departments; I treat them as parts of the same product system.

What I care about

I care about clarity, buildability and taste. I want the product to be understandable without a long explanation, the engineering choices to be grounded in real constraints, and the interface to feel calm, direct and intentional.

I avoid unnecessary complexity. The best version of a product is usually the one where the hard decisions are visible in the simplicity of the final system.

Current focus — mo:ai

I am Technical Founder at mo:ai, building mo: a personal-memory companion that connects software behavior with a physical character. My work brings product decisions, application flows and device integration into the same development loop.

The starting point is a connected prototype supported by a host. An independent device running local AI is a later goal. Interface demonstrations and synthetic evaluations help develop the system, but they do not prove autonomous hardware, battery life or everyday usefulness.

The question I want to answer is practical: can a companion recall something with its source, accept a correction and let someone remove it reliably? That question gives the memory model, permissions and device interaction a common purpose.

Decisions and how I work

In Purge One, making the spool cradle part of the airflow path connected mechanical form to the measurement setup. In LinkagePy, separating numerical computation from the interface loop helped preserve user control during expensive work.

For Scrill, a small menu-bar panel kept the product focused on macOS behavior. In 2yaka, consistency across driver, taxi-stand and administrator views matters more than polishing an isolated screen.

I use AI agents for research, implementation and review. I retain responsibility for scope, design decisions, debugging, the meaning of measurements and delivery. A build is not a physical-device test, and a demonstration is not a user outcome.

When collaborating, I start with context, constraints, existing artifacts and acceptance criteria. I build a small working loop, make failures visible and leave unsupported results open.