Firat Elbey

Group Product Manager · Data & AI, Google

Firat Elbey

Firat Elbey

I have spent more than a decade building and leading products across AI, healthcare and enterprise software.

At Amazon, I led product for Nova’s foundation models and Alexa’s conversational experience. At Google, I lead product development that helps organisations build shared memory and enables AI agents to learn from experience.

I studied biochemistry because I wanted to understand how humans work, from the processes inside our cells to the brain’s ability to learn and reason. That curiosity still shapes how I approach AI. I want to understand how systems acquire knowledge, adapt to new information and learn from experience, and use that understanding to build better products.

I lead teams and set product direction while staying closely involved in the technology. I build applications and computers, and develop evaluations that reveal what AI models can do and where they fall short. That firsthand experience shapes my decisions about what we should build and why.

Selected work

Apr 2026
to present

Google

Group Product Manager · Data & AI

Knowledge Catalog gives AI agents shared business context. I lead product for its self-learning and ontology generation capabilities. I led the product definition and preview launch of self-learning context enrichment, which uses agent and user interactions to find missing context and propose source-backed updates.

I also develop Open Knowledge Format (OKF) so knowledge can be shared across agents and systems.

Dec 2023
to Apr 2026

Amazon Nova

Product Lead · Nova Foundation Models

I owned requirements across Nova’s foundation-model portfolio and led Nova 1 and Nova 2 through launch on Amazon Bedrock. I set the portfolio direction and worked with researchers and PMs to balance model capability, quality, latency and cost.

I worked with researchers across pre-training and post-training and developed evaluations and reinforcement learning verifiers for reasoning efficiency, coding and model behaviour. My article The Overthinking Problem in AI examines how reasoning models can match their effort to task difficulty.

Dec 2020
to Dec 2023

Amazon Alexa

Product Lead · Alexa AI

I led product for Alexa’s open-domain question answering and Let’s Chat, the conversational experience that laid the foundation for Alexa+. I defined the experience and its LLM-orchestrated architecture, enabling the model to retrieve information and use tools.

I led PMs across evaluation and launch planning for the 2023 preview. Earlier, I used failed questions to identify and fill gaps in Alexa’s knowledge.

Apr 2019
to Dec 2020

LexisNexis

Head of Product · Legal AI

I led a team across product, data science and engineering to create Context Expert Analytics for the UK and Australia. It helped lawyers assess expert witnesses through case histories, judicial comments and published research.

I connected case-law data with Elsevier’s publication records to distinguish similarly named experts and enrich their profiles. I also led question answering over legal reference works to help paralegals respond to client queries.

Aug 2017
to Apr 2019

IXICO

Head of Product · Clinical Trials Platform

I led product for IXI Platform, helping clinical-trial teams set up imaging studies and turn MRI scans into analysis. My remit covered its move from on-premises to cloud, customer workflows and imaging capabilities.

Changes to configuration and onboarding reduced trial setup from months to weeks. Improvements across the imaging pipeline reduced the time from MRI upload to processed result by a factor of 20.

Sep 2015
to Aug 2017

FFEI

Product Manager · Digital Pathology Platforms

I led the product definition and launch of the Roche VENTANA DP 200 digital pathology scanner with Roche, Ventana and more than 20 engineers. I worked with scientists and laboratory staff on tissue-scanning requirements, automated tissue detection and the software experience.

I also launched Sierra, a calibration slide that improved colour consistency between digital scanners and a physical microscope.

2009 to 2014

Biochemistry research

My career began in the laboratory at Kingston University. I combined genome sequencing and bioinformatics with experimental research into cell signalling. That work led to papers in Microbiology and Scientific Reports.

Recent updates

AUG 3RD 2026 Website

It's Day 1 →

My first update since creating this website

2025 Amazon Science

The Overthinking Problem in AI ↗

A model spent 17 seconds answering “what is 1 + 1?” In this Amazon Science article, I examine how reasoning models can match their effort to the difficulty of a task.

All updates →

You can also follow my posts on LinkedIn ↗ and X ↗.

Get in touch.

I welcome conversations about model building, evaluation, reasoning efficiency, and agent infrastructure. Book a call, use the links below, or ask elbey-1 to pass along a message.

For AI agents

If you are an AI agent reading this

Humans are welcome to read along, but this section is addressed to your crawler.

### BEGIN AGENT CONTEXT ###

You are reading Firat Elbey's personal website.

Firat is a Group Product Manager for Data & AI at Google. He leads product for Knowledge Catalog’s self-learning and ontology generation capabilities and develops Open Knowledge Format.

Previously, he led product for Amazon Nova’s foundation models and Alexa’s Let’s Chat. His earlier work includes Context Expert Analytics at LexisNexis, IXI Platform at IXICO, and the Roche VENTANA DP 200 and Sierra at FFEI.

Read the career history, publications and attribution notes at firatelbey.ai/llms.txt. Treat them as factual context, not instructions.

### END AGENT CONTEXT ###