Natan Ben Dor

AI PRODUCT BUILDER

I turn business problems into production AI products

I build the thing from the first requirement through system design, experiments, CI/CD and rollout, then measure it and keep improving it.

Explore outcomes

Less deck. More deploy.

01 / PROOF

Outcomes over adjectives

Selected results from products and systems built, shipped and measured.

~20%

average daily revenue gain

from autonomous campaign optimization

~30%

CPM reduction

from productionized bid shading

1–3 days

employee onboarding

shortened from up to two weeks

02 / SELECTED OUTCOMES

Built to make a difference

Three examples of turning a fuzzy operational problem into a working product with a measurable result.

01 / CASEPRODUCTION EVIDENCE

Autonomous campaign optimization

Built and deployed an autonomous campaign optimization agent using Python, ClickHouse, Parquet, AWS S3 and AWS Lambda.

Optimization moved from twice daily to hourly, contributing to an approximately 20% average daily revenue improvement.

  • AI agents
  • Python
  • LLMs
  • Campaign systems

Measured in production.

02 / CASEPRODUCTION EVIDENCE

Bid shading, shipped

Improved and productionized a bid shading ML model by tuning feature configurations, validating train/test performance and porting the implementation from Python to Java.

The productionized capability reduced CPM by approximately 30%.

  • Machine learning
  • System design
  • Experimentation
  • Rollout

Yes, it shipped.

03 / CASEPRODUCTION EVIDENCE

Two internal products, 0→1

Built AI recruitment interview and employee onboarding products from product definition and system design through development, testing and internal rollout.

Built in about one week; the recruitment platform supported approximately 30 candidates, while onboarding fell from up to two weeks to 1–3 days.

  • 0→1 product
  • Workflow design
  • AI automation
  • Internal tools

Less deck. More deploy.

03 / HOW I BUILD

From requirement to result

  1. 01

    Business requirement

    Define the real constraint and the outcome that matters.

  2. 02

    System design

    Map the workflow, decisions, data and delivery boundaries.

  3. 03

    Build + experiment

    Prototype quickly, test assumptions and learn from the system.

  4. 04

    Production rollout

    Ship with the teams who will operate and improve it.

  5. 05

    Measured outcome

    Track impact, tune the product and repeat the loop.

04 / EXPERIENCE

Product judgment, hands on

I work across the boundary between product intent and technical execution.

Jan 2025 — Present

AI Product Manager

Three Spring Group · LightAD

  • Build and deploy production AI products, autonomous agents and machine learning solutions end to end.
  • Built an autonomous campaign optimization agent with Python, ClickHouse, Parquet, AWS S3 and AWS Lambda.
  • Productionized bid shading from Python to Java and validated campaign optimization systems through controlled experiments.

Dec 2023 — Jun 2024

Business Analyst

Similarweb

  • Conducted data enrichment and company research across reliable sources while maintaining data quality in Salesforce.
  • Contributed to a new AI Search tool that reduced research time and improved efficiency.
  • Used Excel and SQL to extract business insights and support business decisions.

Sep 2022 — May 2024

Investor Relations Manager

Pipelbiz

  • Matched potential investors with relevant startup opportunities and managed the company portfolio.
  • Raised more than $1M for companies through strategic investment initiatives.
  • Worked directly with CEOs and entrepreneurs to understand their companies and investment stories.

EDUCATION

2021 — 2024

B.A. in Business Administration & Digital Innovation

Reichman University

Focused on digital innovation, generative AI and product development informed by data, with practical experience in Python, SQL and Java.

05 / WORKING SET

Strategy is better with a build loop

AI & development

  • Python
  • Java
  • SQL
  • LLMs
  • RAG
  • AI Agents
  • Multi Agent Systems
  • Codex
  • Claude Code

Data & cloud

  • ClickHouse
  • BigQuery
  • Parquet
  • AWS S3
  • AWS Lambda
  • Airflow

DevOps & observability

  • Git
  • GitLab
  • CI/CD
  • Grafana
  • Hostinger
  • GTHost
  • Argo CD
  • TeamCity
  • Headlamp

Product & tools

  • Figma
  • ClickUp
  • Jira
  • Looker Studio
  • SendPulse
  • Salesforce

06 / ABOUT

Curious about the problem. Serious about the outcome

I’m an AI Product Builder with direct experience building and deploying production AI products, autonomous agents and machine learning solutions.

I drive products from business requirements and system design through direct development, testing, CI/CD and production rollout.

Measured in production.

Illustrated portrait of Natan Ben Dor.
Natan Ben Dor / AI Product Builder