Plan
Requirements, constraints, and trade-offs are explored with an AI assistant before code exists. Assumptions are written down, not carried in my head.
invent.ai · Senior/Staff Software Engineer
Implementation moves to AI agents. Context and guardrails stay with me.
I am a Senior Product Developer based in İzmir with more than six years of shipping web products, mostly on the frontend side of API-driven systems. I run an AI-native workflow on Claude Code, custom skills, and MCP connections to production tooling, and I am building systems that hand implementation to AI agents end to end while specification, review, and verification stay with me. I am applying at the Senior band and naming backend service ownership as my growth edge rather than a claim.
- More than six years shipping web products across React, Next.js, Vue, and Angular.
- Building systems that delegate implementation to AI agents, with tests and CI as the gate.
- Deliberate context management: lean AGENTS.md, scoped agent rules, data loaded on demand.
- Clear about the growth edge: production Express or Nest service ownership.
End-to-end AI leverage
The assistant is in the whole lifecycle, and so is the review.
Planning, implementation, verification, and operations each get AI leverage. Each also keeps a human accountable for the outcome.
Build
AI agents implement and write tests against an explicit spec. Every diff is reviewed as if a colleague wrote it, and simpler options are preferred.
Verify
Vitest and Playwright run in CI. Typecheck, lint, and tests are the guardrail that lets AI-generated code move quickly without hiding regressions.
Operate
Sentry and Jira are wired into the assistant through MCP and Graylog through a custom skill, so investigation, incident notes, and follow-up tickets start from real production data.
Evidence
How I already work, and the products that shaped it
AI-native workflow · daily practice
This is not a tool I tried once. The assistant sits in planning, implementation, debugging, incident investigation, and pull requests. I am now taking it a step further: a process that delegates implementation to AI agents end to end, automates the repeatable steps, and keeps the guardrails inside the same system.
- Project instructions in AGENTS.md and agent rules, kept short and scoped, that bias toward small, verifiable changes.
- Custom Claude Code skills and slash commands for Jira ticket context, Sentry incident reports, Graylog user-journey reconstruction, code review, commits, and PR creation.
- MCP servers connecting the assistant to Jira, Sentry, and an internal product data service.
- Context treated as a limited resource: ticket, incident, and product data are pulled in on demand through skills and MCP instead of pasted.
- This application site itself: a typed Next.js codebase with Vitest coverage, built and maintained through that workflow.
eSIM.io · Teknasyon
VisitAs Senior Product Developer, my current product is a B2C eSIM purchase journey built with Next.js and TypeScript. It is where my AI-assisted workflow, API collaboration with backend teams, and production quality habits meet.
- Built lightweight Next.js endpoints and contributed CRUD endpoints to backend projects.
- Worked with backend engineers on API contracts and on where frontend-facing logic should live.
- Vitest and Playwright tests written in an AI-assisted TDD loop and executed in CI.
- Shipped A/B page variants, collaborated with analytics, and contributed to SEO and PageSpeed.
Nesine.com
VisitA high-traffic React product in a frontend monorepo during a migration from legacy .NET surfaces to a React SPA. Performance regressions there were user-facing production problems.
- Investigated rendering with React Profiler and component update analysis.
- Applied virtual-list and re-render optimisation in production surfaces.
- Shipped within established monorepo conventions and shared patterns.
- Treated rendering, loading, and regressions as production health signals.
Datapad
A data-focused startup building KPI dashboards from multiple sources. A small team meant I contributed briefly on the backend side as well as owning frontend delivery.
- Connected multiple data sources to user-facing KPI surfaces with Next.js and TypeScript.
- Contributed to backend work alongside frontend delivery in a small team.
- Experienced fast iteration, uncertainty, and the cost of unverified assumptions.
Rockads · Teknasyon and Petal Maps Studio · Huawei
Two integration-heavy products: an advertising-management centre built on third-party platform APIs, and an enterprise map-customisation tool inside a large international codebase.
- Integrated external advertising platform APIs and chart-heavy dashboards through a Vue 2 to Vue 3 migration.
- Handled complex state between UI and a map engine, with perceived-performance constraints.
- Worked in an international team with formal review and approval processes.
Role fit
Where the posting and my evidence meet, and where they do not yet
Positioning principle
More than six years of shipping product and an AI-native workflow are the evidence. Backend service ownership is named as the growth edge, not dressed up as history.
Contemporary JavaScript and TypeScript
More than six years of ES6+ and TypeScript across React, Next.js, Vue, and Angular. Asynchronous data flows, caching, state ownership, and framework migrations are everyday work, not portfolio exposure.
End-to-end AI leverage and reusable tooling
I use Claude Code across planning, implementation, debugging, incident investigation, and PR workflows, and I am building a process that delegates implementation to AI agents end to end. I maintain AGENTS.md files, agent rules, custom skills, slash commands, and MCP connections that compound over time, and I keep context lean so each task starts from relevant information only. I also know when a change is faster or safer by hand.
AI guardrails and system health
AI-generated code goes through the same gate as mine: Vitest and Playwright tests in CI, typecheck, lint, and human review of every diff. Sentry and Graylog are connected to the assistant, so debugging starts from production data. I have not owned rollout or rollback infrastructure for backend services; I have worked inside teams that did.
APIs and the data flow to the frontend
The posting asks the engineer to work with frontend developers on well-designed data flow. I have been on the other side of that conversation for more than six years, built lightweight Next.js endpoints, and contributed CRUD endpoints to backend projects. I know what a good contract looks like from the consumer's seat.
Code review, mentorship, communication
Code review and cross-team communication are part of my current role, and public talks on web performance, Next.js, and JavaScript internals show I can make a technical trade-off understandable. English is comfortable for interviews and daily remote work.
Honest growth edge: backend service ownership
I do not have two years of production Express, Hapi, Sails, or Nest experience, and I have not owned microservice architecture or database operations. I am naming this directly so we can discuss how invent.ai onboards engineers into its data-intensive backend and how fast that ownership can realistically grow.
invent.ai
My reading of the company and the role
Public-source research kept separate from claims about invent.ai's internal architecture, roadmap, or team decisions.
01
invent.ai sells an AI-decisioning platform for retailers with Planning, Inventory, and Pricing modules, and its Remi agent monitors demand and stock at store and SKU level.
The product output is a decision, not a page. Services carry forecasts and recommendations at SKU, store, and day granularity, so API correctness, performance, and clear data contracts are product features rather than internal details.
AI-Decisioning Platform overview02
In September 2026 invent.ai published a Remi MCP server so AI clients such as Claude can query inventory, sales, and forecast data with tenant isolation.
The company builds with the Model Context Protocol on the product side, not only in developer tooling. I use MCP servers daily for issue tracking, error monitoring, and an internal data service, so the pattern and its trust boundaries are familiar territory.
Remi MCP announcement03
Across its current openings the stack is polyglot: a Node.js and Express API layer with PostgreSQL, MongoDB, Redis, and Elasticsearch, a Python and PySpark data layer on AWS Lambda, and React frontends.
The API layer sits between batch data pipelines and React interfaces. That makes the posting's line about designing data flow with frontend developers central to the role, and it is where more than six years on the consuming side of those contracts is directly useful.
Senior/Staff Software Engineer posting04
The posting's AI Proficiency section is unusually specific: CLAUDE.md files, agent rules, skills, slash commands, internal MCPs, and judgment about when not to use AI, paired with testing, CI, safe rollout, and observability.
That is a description of how I already work, including the judgment part. Pairing AI speed with guardrails tells me the team wants trustworthy AI-assisted engineering rather than speed alone, which matches the working rules I apply today.
Senior/Staff Software Engineer posting05
Founded in 2013 by Prof. Gurhan Kok and formerly named Invent Analytics, the company closed a Series B in 2024 and states 'Science at our core' and 'Collaboration over hierarchy' as values, with remote teams in Istanbul, Philadelphia, Amsterdam, and London.
A science-led company with academic advisors argues engineering decisions with evidence, which is the communication style I prefer. Remote work from İzmir with European and US overlap is my normal operating mode.
Join invent.ai06
Enterprise retail customers such as Alo Yoga, Boots UK, Mavi, and Five Below, a 2026 Retail Systems 'AI Project of the Year' award, and SOC 2 and ISO 27001 coverage are public signals.
Enterprise customers make the 'system health' half of the posting real: rollback safety, observability, and security are commercial requirements. This is the part of the role I would be growing into fastest, and I would rather say so than pretend otherwise.
Retail Systems Awards 2026Communication proof
Technical depth is more useful when it can be shared clearly
I have given public talks on web performance, Next.js, JavaScript internals, and data structures. They are practical evidence of preparation, technical curiosity, and explaining trade-offs to an audience.