Kish Dubey

I'm Kish
Building AI-native systems for commercial real estate 🏢

Get in touch →

Background

I build AI-native platforms that automate complex decision workflows in commercial real estate, turning weeks of underwriting and due diligence into minutes.

Currently Founder in Residence at Antler, focused on the intersection of AI, automation, and real estate finance. Previously built ML systems at IBM, shipped product at YC-backed Kaagaz (10M+ users), and bootstrapped a B2B SaaS to five figures before hitting a wall and learning what not to do.

I'm interested in applying AI to solve real problems in industries that have been slow to adopt it. Commercial real estate is one of the last trillion-dollar industries still running on spreadsheets and gut instinct, that's the opportunity.

Based in Toronto. Always happy to discuss AI in real estate, automation workflows, or startup ideas.

Experience

Stealth CRE Tech Startup

2025 - Present

Founder, CTO

Building an AI-native platform that automates complex decision workflows in commercial real estate, turning weeks of underwriting, analysis, and due diligence into minutes. Focused on the Canadian CRE market with tools that make data-driven investment decisions faster and smarter.

Antler

2025 - Present

Founder in Residence

Part of Antler's intensive residency program (< 10% acceptance rate). Building and validating AI-native solutions in commercial real estate, backed by a global VC that's funded 900+ companies across six continents.

IBM

2023 - 2025

Software Engineer, AI/ML

Built scalable backend and frontend systems for enterprise data monitoring, compliance automation, and analytics visualization. Shipped production ML models and cloud-native applications used by Fortune 500 clients.

Pinkbot

2022 - 2023

Software Engineer

Built mapping and geospatial systems for autonomous delivery robotics. Worked on real-time path planning and environmental modeling.

IBM

2022 - 2022

Data Scientist

Built a Graph2Seq-powered ML system that improved SQL-to-ticket matching accuracy by 75% (to 79.4%), presented to 450+ stakeholders. Part of IBM's elite Extreme Blue technical leadership program.

bordr

2021 - 2021

Founder

Bootstrapped a B2B SaaS to 9 clients and five-figure revenue. Hit growth constraints around distribution and customer acquisition, learned more from this than any success.

Kaagaz (YC W'22)

2020 - 2020

Software Engineer

Early engineer at YC-backed document scanning startup. Contributed to product growth from early stage to 10M+ downloads and 4.8 rating.

EdCast by Cornerstone

2020 - 2020

Machine Learning Engineer

Developed an AI-powered, machine-learning-driven walkthrough system that intelligently adapts to UI changes, ensuring seamless user onboarding despite dynamic interface updates.

Bruyère

2019 - 2019

Software Engineer

Incubated SmartCue, a smart home system for dementia care that helps patients follow daily routines. Tested with elderly patients at Bruyère Hospital under Dr. Frank Knoefel and Professor Bruce Wallace.

Writing

Two Weeks, Three AI Agents, and One Knowledge Graph: Building a Research Assistant

June 18, 2025

"Over the past two weeks, I built a personal research assistant that combines knowledge graphs, AI agents, and semantic search, designed to help analyze academic papers and uncover research gaps. Turns out coordinating AI agents is like herding cats, but smarter cats that occasionally refuse to talk to each other..."

Read full article on LinkedIn →

Prompt Engineering for Document Analysis: What I Learned Moving from GPT-4 to Claude 4 🧠

May 28, 2025

"The Problem: I've been working on structured data extraction from messy financial documents, think scanned invoices with handwritten notes, mixed-language contracts, tables spanning multiple pages. My GPT-4 pipeline was hitting a wall at around 60% accuracy for the test set of 500 documents that were manually labeled..."

Read full article on LinkedIn →