— İzmir, Türkiye — Portfolio

Gülşah Küçükcan

Project Manager · Business Analyst · AI Product Discovery

İzmir, Türkiye·10+ years in software delivery·gulsah.kucukcan@gmail.com·LinkedIn

I turn ambiguous problems into software teams can actually build — and, increasingly, into AI-powered products worth building. Ten-plus years across healthcare, retail and enterprise software have taught me that the hard part is rarely the technology; it's getting the right people aligned on the right problem.

01 — About

A Computer Engineer who moved into the room where decisions get made

I'm a Project Manager and Business Analyst with over ten years in software and technology, including seven-plus years leading end-to-end delivery for cross-functional teams. My background is in Computer Engineering — a foundation I use daily, not to write production code, but to ask sharper questions, evaluate trade-offs honestly, and work as a genuine peer to engineering rather than a translator standing outside it.

I've spent my career at the point where business needs, product decisions and engineering constraints meet — across healthcare, retail and enterprise software, and more recently, AI-powered products. What stays constant is the work itself: understanding a problem well enough to scope it properly, aligning the people who need to agree on a direction, and staying close enough to delivery that scope, risk and quality don't quietly drift.

Outside of AI product work, I care a lot about how teams function as teams — which is part of why this site has a whole section on that, not just a list of tools I know.

Project & Delivery

Project ManagementAgile / ScrumEnd-to-End DeliveryRisk & Dependency ManagementResource & Task Planning

Business & Product

Business AnalysisRequirements EngineeringUser Stories & Acceptance CriteriaProcess MappingProduct StrategyKPI Definition

Stakeholder Management

Client ManagementExpectation ManagementCross-functional CollaborationPresentation & Reporting

Technical

SDLCAPI IntegrationAI & Data ProjectsSQLPower BIUAT & Regression Testing

Career at a glance

2025 — Present
Project Manager & Business Analyst — AI-powered healthcare analytics platform, U.S. hospital network
2020 — 2025
Project Manager & Business Analyst — Healthcare applications suite, U.S. clinicians & home health aides
2019 — 2020
QA Specialist — Biographical database platform, major Turkish university
2018 — 2019
Brand & Product Manager — E-commerce platform launch, agricultural products company
2017
Business Analyst — Operational BI & workflow optimization, insurance company
2016
Software Developer — Hyperbaric chamber control software, industrial systems company
2015
Business Analyst & Software Tester — Heavy equipment simulation software, learning technology company

02 — General Projects

Delivery work across healthcare, retail and industrial software

A selection of the software delivery work that shaped how I manage projects today. Company names are withheld at my employers' and clients' request — the work, my role, and my responsibilities are described exactly as they were.

Aug 2020 – Aug 2025Healthcare software client, U.S.

Healthcare Applications Suite

Managed an 8-member cross-functional team delivering and maintaining five integrated healthcare applications used by doctors and healthcare aides across the U.S.

  • Owned end-to-end project activities — planning, prioritization, task allocation and delivery tracking — within an Agile environment
  • Translated client requirements into technical documentation, user stories and acceptance criteria
  • Coordinated maintenance, enhancement and API integration work across all five applications
  • Led daily Scrum meetings and kept engineering, clients and business stakeholders aligned
Jul 2022 – Jun 2023Retail client, U.S.

Warehouse Management System

Managed the full delivery lifecycle of an Android-based warehouse management system for handheld devices.

  • Defined functional specifications for order packing, inter-store transfers and last-mile delivery
  • Documented end-to-end workflows and use cases against real warehouse operations
  • Planned and coordinated Agile/Scrum delivery across developers, testers and stakeholders
  • Conducted manual and regression testing; coordinated deployment and post-launch fixes
Sep 2019 – Aug 2020Major Turkish university

Biographical Database Platform

Supported development of a web application managing a dynamic database of Turkish-language biographies dating back to the 1800s.

  • Gathered and translated user and client requirements into functional specifications
  • Prepared use cases, workflow diagrams and testing documentation
  • Conducted functional testing and validated database integrity across authors, publications and biographical records
Jan 2018 – Aug 2019Agricultural products company

E-Commerce Platform Launch

Managed the launch and end-to-end operation of an e-commerce platform, including website setup, payment and cargo integrations.

  • Built SEO and digital marketing initiatives to grow brand visibility and online sales
  • Ran market and competitor analysis to shape go-to-market strategy
  • Optimized the online customer journey with suppliers, logistics and agency partners
Jan 2017 – Sep 2017Insurance company

Operational BI & Workflow Optimization

Conducted business and requirements analysis to identify operational issues and define technology-driven solutions.

  • Extracted and analyzed SQL data; built Power BI dashboards for management decision-making
  • Translated business requirements into functional specifications for the development team
  • Ran User Acceptance Testing for new features and enhancements
Mar 2016 – Nov 2016Industrial control systems company

Hyperbaric Chamber Control Software

Developed and maintained control software for diving and hyperbaric chamber systems.

  • Designed software modules, implemented enhancements and resolved defects
  • Integrated applications with PLC systems using the TwinCAT environment
  • Tested and validated software under varying pressure conditions; wrote technical documentation
Feb 2015 – Nov 2015Simulation & learning technology company

Heavy Equipment Simulation Software

Tested heavy equipment simulation software built with Unity and C#.

  • Designed and executed test cases; documented defects with the development team
  • Gathered and validated client requirements for simulation scenarios and interfaces
  • Prepared test plans, bug reports and release documentation

03 — AI Projects

Building AI-powered products, from problem discovery to execution

My experience with AI sits at the intersection of product management, project management and emerging AI technologies. I've worked on AI-powered products and concepts where my role was to understand the business problem, structure product discovery, evaluate solution alternatives, align stakeholders, and work closely with engineering to turn ideas into feasible solutions.

I'm not an AI engineer — and I don't position myself as one. My strength is understanding the technology well enough to make better product decisions, ask the right questions, and work effectively with technical teams.

ProblemDiscoveryAlternativesTrade-offsBudgetFeasibilityExecutionIteration

I worked on an AI-native operations platform designed to support complex business workflows through AI-powered processes and specialized agents. My role focused on product discovery, project structure, workflow definition and alignment between business needs and technical possibilities.

My role as Product Owner / Project Manager

I translated high-level business needs into something the product and engineering teams could understand and execute — including:

  • Structuring business requirements and defining desired outcomes
  • Breaking complex workflows into manageable workstreams
  • Leading product discovery and prioritization with the customer
  • Exploring where AI could create meaningful value, and evaluating solution approaches
  • Coordinating between business, product and engineering; supporting scope and feasibility calls
  • Weighing budget, complexity and delivery implications; tracking cross-workflow dependencies

The discovery challenge

The real question was never simply "can we use AI here?" It was: which parts of the workflow should be automated, which should be AI-assisted, and where should human decision-making remain?

Single AI agentA general-purpose agent handling the entire workflow.
Multiple specialized agentsDifferent agents responsible for specific tasks.
AI-assisted workflowAI supports specific stages while humans stay responsible for key decisions.

Each approach carried different implications for scope, reliability, engineering effort, operational cost and scalability.

Working with engineering

My role wasn't to define implementation details independently, but to bring product requirements and business context into the technical discussion — what was feasible, what needed new infrastructure, what would add development effort, where orchestration would get unnecessarily complex, and what should stay human-controlled.

Budget & business impact

Budget was part of the decision-making process from discovery onward: model/API usage, engineering effort, infrastructure, maintenance complexity, expected efficiency gains and scalability — weighed against the value the solution would actually create.

This project strengthened my ability to act as the bridge between business requirements, product decisions and technical execution in an AI environment.

I worked on the preliminary product discovery of an AI-powered product discovery system built around five separate databases. The vision: a customer describes what they need once, and the AI finds the relevant product information and manuals across multiple systems. The challenge was making a complex backend feel simple from the user's side.

My role

  • Understood the customer problem and mapped the existing information landscape
  • Defined the desired conversational experience and structured the discovery process
  • Explored agent-based approaches and evaluated alternatives with engineering
  • Weighed implementation complexity and cost; turned the concept into a structured technical direction

Alternatives explored

A — One general-purpose agentSimpler to prototype; risked becoming responsible for too many types of data and tasks.
B — Database-specific agentsEach database gets its own specialist; risked a complicated cross-database routing problem.
C — Central routing + specialized agentsA routing layer interprets intent and dispatches to the right agent(s) — the direction we explored, balancing a simple customer experience with specialized backend responsibilities.

Collaboration with engineering

The architecture was explored together with engineering, not defined purely from the product side — asking what the user needs, what the AI should decide versus the system, and what would be expensive or hard to maintain. My job was connecting those technical constraints back to product priorities and customer value.

Budget & business impact

Integration effort across five databases, AI/API usage, number and complexity of agents, maintenance and scalability — the goal was never the most sophisticated architecture, but the one that delivered the right customer value at a reasonable cost.

This project strengthened my experience in AI product discovery, data-heavy products and cross-functional collaboration with engineering.

I worked on the preliminary product concept of an AI-powered educational assistant. The initial idea was a conversational chatbot, but discovery pushed the experience further: the AI could understand a student's question, identify the relevant topic, direct them to the right learning content, and generate example questions for practice.

My role

  • Understood learner needs and defined the chatbot's role
  • Mapped questions to learning topics and designed the conversational flow
  • Defined user journeys and explored content-navigation and AI-generated practice
  • Worked with technical stakeholders on feasibility, scope and priorities

Alternatives considered

Answer botThe chatbot simply answers the student's question.
Content navigatorIdentifies the relevant topic and redirects the learner to existing content.
Interactive learning assistantAnswers, identifies the topic, redirects to content and generates follow-up questions — more value, but more AI usage and complexity.

Budget & business impact

The key question: can AI meaningfully improve engagement and personalization without disproportionate operational cost? That framed the MVP around the features most likely to create value, rather than automating the whole learning experience from day one.

This project gave me experience in conversational product design, AI-assisted personalization and defining an AI MVP together with technical teams.
Explore Tinvie → tinvie.com

Tinvie is different from the projects above: in earlier work I contributed to AI product discovery within teams. With Tinvie, I'm building the product myself — an anonymous emotional sharing platform built around the idea that identity should matter less than how you feel.

My role

As the product owner of my own product, I'm involved across the entire lifecycle: discovery, strategy, user research and problem definition, feature prioritization, UX, AI feature definition, technical architecture discussions, development coordination, testing and iteration — experiencing responsibilities normally distributed across Product, Project Management, Engineering and QA.

Where AI comes in

AI-powered content moderation, speech-to-text, translation, audio/content filtering and AI-assisted safety workflows. One example: processing original user audio through Recording → Speech-to-Text → Moderation → Translation/Processing → Delivery.

Product vs. technical decisions

Should audio be processed immediately? Should moderation happen before or after speech-to-text? Should translation happen before publishing? Should every piece of content go through every AI service? Each choice trades off user experience, safety, performance, AI cost and infrastructure cost.

Budget & MVP thinking

As an independent product, Tinvie makes the relationship between feature value and AI cost especially visible. Every new AI capability gets weighed against API/model cost, infrastructure, development effort, maintenance, expected user value and MVP scope — not just "can we build this?" but "is it valuable enough to build now?"

Tinvie is where my interest in AI, product management and building products come together most directly.

How I work on AI products

Across these projects, I see my role as a bridge between product, business and technology. I don't need to write the underlying AI infrastructure myself to contribute meaningfully — I need to understand enough about the technology to ask the right questions, challenge assumptions, evaluate alternatives, discuss feasibility with engineers, understand cost implications, and translate business needs into clear requirements that keep teams aligned around the outcome.

01Understand the problemWhat are we actually trying to solve?
02DiscoveryWhat do users, the business and existing systems tell us?
03Explore alternativesWhat are the possible ways of solving it?
04Evaluate trade-offsImplications for UX, technology, cost, timeline and scalability.
05Collaborate with engineeringWhat's technically feasible, and what should we build first?
06Define & prioritizeWhat creates enough value to justify its cost and complexity?
07ExecuteTurn the product direction into actionable work for the team.
08Learn & iterateUse feedback and real-world results to improve the product.

I don't start with "where can we add AI?" I start with "what problem are we solving, what are our options, and where can AI create meaningful value?" Effective AI product management is about finding the balance between user value, business value, technical feasibility, cost, complexity and speed — that's where I believe the strongest AI products get built.

04 — Team Spirit

Delivery is a people problem before it's a process problem

Every project I've managed has taught me the same thing in a different disguise: the roadmap rarely fails because of a missing feature — it fails because people didn't really understand each other. Two developers reading the same ticket differently. A client and an engineer meaning different things by "urgent." A tester who doesn't feel safe flagging a risk out loud. None of that shows up on a Gantt chart, and all of it is where projects actually slip.

I take team health as seriously as I take scope and timeline, because I've seen the two are the same problem. A team that genuinely knows how its members work — what makes someone go quiet when they're stuck, who needs to think out loud, who needs quiet to think at all — moves faster with less friction than a team that is simply "aligned" on paper. That kind of trust isn't a soft add-on to project management; it's the thing that makes estimates honest, retrospectives useful, and hard conversations possible before they turn into missed deadlines.

This isn't about forcing fun. It's about building small, repeatable moments where people get to know each other as people, not just as tickets in a sprint — and protecting those moments even when the schedule gets tight, because that's exactly when they matter most.

1:1s that aren't status reports

A recurring check-in where the first question is "how are you, actually" — not "where's the ticket." Status has a stand-up for that; the 1:1 is for energy, blockers, and what's quietly wearing someone down.

A real "how I work" onboarding

New teammates get more than tool access — a short, honest exchange about working styles, communication preferences and what motivates each person, so the team isn't guessing for the first three months.

Retros that look at the team, not just the backlog

Alongside "what slowed delivery," I ask what slowed us as a team — where trust wobbled, where someone felt unheard. Process debt and people debt get tracked the same way.

Cross-functional show-and-tells

Engineers, QA, business and design walk each other through their own work directly, in their own words — so respect is built on seeing the work, not just hearing a summary of it.

No-agenda time, protected on purpose

A recurring slot with nothing to deliver — coffee, a walk, a shared lunch. It's the first thing that gets cut under pressure, and exactly the thing that pays off when pressure hits.

Wins named out loud

Not just launches — the unglamorous saves: the bug caught before it shipped, the risk someone flagged early enough to matter. Naming those keeps people willing to speak up next time.

05 — Background

Education & certifications

B.Sc. in Computer Engineering — İstanbul Bilgi University
2010 – 2014
Professional Scrum Master I (PSM I) — Scrum.org
Project Management Certificate — Bahçeşehir University
2020
Principles of Artificial Intelligence — PMI.org
2022

Languages & additional expertise

English — C1Artificial IntelligenceKPI & Data AnalyticsSoftware TestingDigital Project Management