Menu
About Focus Work Build Contact

Tümay Turhan

I begin with the problem, not the model.

Independent researcher, builder, systems thinker. I design applied AI where people, capital, and constraints intersect.

Agentic systems, context, and evaluation—where risk and attention are part of the spec. Open to aligned collaboration.

Selected work GitHub

Intersection

My path is not a straight line—it is an intersection. Civil engineering and construction management taught me loads, failure modes, and accountability when things break. An MBA, project finance, and years in underwriting taught me how money moves, who actually decides, and how to reason under incomplete information.

I begin with the problem, not the model. I treat people, incentives, risk and system shape as the frame—and ask where models earn their place.

A belief I return to: to understand AI in the wild, you start with people. Then money. Only then do you see where augmentation fits—without illusion.

I learn by building; I build with intention.

Domains

What I focus on

Agentic Systems

Build systems when one model is not enough.

Routing, multi-agent flows, and decision chains that hold under real conditions.

Context Engineering

The bottleneck is rarely the model. It is context.

What you give, when you give it, and how decisions are framed.

Evaluation & Reliability

Evaluation is architecture, not a final step.

If you cannot measure behavior, you do not know what you built.

AI Under Constraints

Constraints are the design.

Latency, cost, and human attention shape what works.

Product-Oriented AI

If it is not usable, it does not matter.

People must be able to rely on it.

Portfolio

Selected work

Case-shaped work. Problem, approach, and the thinking behind each decision.

YouTube comment analytics — dashboard overview
YouTube comment analytics — topics, intent, sentiment and comment list

YouTube comment intelligence

Analytics

Problem

Comment threads hold product and sentiment signal; raw text does not scale to product or strategy decisions.

Why it mattered

Without a pipeline and a stable ontology, every analysis reinvents categories—BI and stakeholders never get comparable views over time.

Approach

Ingest via YouTube API into layered BigQuery tables, LLM enrichment for structured attributes, Streamlit for exploration and Looker Studio for reporting.

Key insight

Unstructured comments become useful only when mapped into consistent sentiment and topic signals—repeatable, decision-grade views for creators and sponsors.

Repository Looker Studio
Steam 2025 — executive summary visual
Steam — Looker Studio dashboard KPIs and market trend
Steam analysis — chart from README
Steam analysis — chart from README

Steam marketplace analytics & forecasting

240k+ games

Problem

The market is large, but signal is weak—price and performance drivers hide behind noise and uneven data quality.

Why it mattered

Without an honest data layer, forecasts mislead. Weak signals must surface, not disappear, if decisions are to be trusted.

Approach

BigQuery + dbt, Prophet for growth, gradient boosting / RF for price drivers, Looker Studio for exploration.

Key insight

Not all signals deserve attention—Metacritic vs price is weak (r ≈ 0.23). Removing noise is as valuable as modeling signal.

Repository Looker Studio
Olist BigQuery SQL case study — screenshot 1
Olist BigQuery SQL case study — screenshot 2
Olist CEO management dashboard — satisfaction drivers
Olist CEO management dashboard — financial summary and waterfall
Olist CEO management dashboard — seller portfolio optimization

Brazilian e-commerce analytics

100K+ orders

Problem

Public e-commerce data is easy to quote and hard to trust—analysts need clear numbers; leaders need profit context, not adjectives.

Why it mattered

One story, two surfaces: SQL work for rigor, then a small management dashboard anyone can read.

Approach

SQL project: clean joins, headline KPIs (~15.8M BRL revenue, high on-time delivery, ~6% repeat buyers). Dash: satisfaction translated into money on the page.

Key insight

Operations look healthy; loyalty does not—the dashboard's job is to show that gap in plain money terms.

Repository — SQL Repository — CEO Dash
Data Mesa — focus timer, forest ambience, task queue, and reflective quote
Data Mesa — break mode, space scene, lofi player, and LIFE task list
Data Mesa — focus session with night-rain scene and ambient controls

Data Mesa — calm focus workspace

Product

Problem

Most focus tools demand attention—features, tracking, and pressure fragment attention further.

Why it mattered

Focus follows environment and cognitive load; without a calm, predictable space, attention destabilizes.

Approach

Ambient-first workspace: minimal UI, controlled sensory input, familiar scenes (rain, night, forest). Next.js, TypeScript, CSS Modules; bilingual UX; Vercel. State stays local—your sessions stay yours.

Key insight

Focus is regulation, not timers—environment shapes attention. Shipped as thinkslow.vercel.app.

Live app Repository
1 / 5

Build

Ecosystem

Live products show how ideas behave.
Technical work shows how they are built.

Data Masası

youtube ↗

Structured tutorials on SQL, BigQuery, and analytics—teaching as a slow, quiet surface.