Screen CVs with AI.
Interview Smarter.
A practical guide for using an LLM to shortlist faster, spot candidate patterns, and know what to probe in the first 15 minutes.
Field notes, not science. Use it as a first-pass lens, never as a hiring verdict.
Selected work
Selected cases show what was built, what broke, and what transferred to the next bet.
cheap.market
Cross-border commerce experiment: China supply into affordable, social-first shopping. Marketplace mechanics, social discovery, AI-assisted catalog ops, logistics, payments, and real unit economics.
Read the investor deck ->Hypee
Consumer photo & video editor. Acquired by Yandex for ~$3.1M and integrated into Yandex Zen.
Read more ->AI Boost
AI photo & video editor. Top App Store rank in Brazil at ~16K new users/day, profitable on Meta ads, and exited for ~$450K.
Read the story ->10+ consumer products
7 years across marketplaces, AI apps, creator tools, mobile growth, serverless infrastructure, and real-world operations. Some worked, some didn't. All of them taught something.
TikTok CPI $0.09, paid CAC $8
Problem: The consumer commerce funnel had thin margins, and standard CAC math was misleading.
Result: Live commerce experiments reached $0.09 TikTok CPI and $8 paid CAC per buyer, with documented unit economics.
Stable Diffusion API in 2 days
Problem: Competitor shipped an AI image product; we had no backend and two weeks felt too long.
Result: Production API live in 48 hours on serverless AWS. The hard part wasn't the build. It was everything after launch.
Serverless at founder scale
Problem: Tiny teams can't carry heavy infra across 10+ products.
Result: Repeatable Lambda/DynamoDB/S3 stack from Hypee through cheap.market. I still write the production code myself.
The Hunt
Between building companies, I do one thing obsessively: I take the whole venture market apart to find where the next one hides.
Every month I map every funding round I can find, by stage, business model, monetization, capital intensity, and geography. Not to report the news. To answer one question for myself: where is the money going, where is it not going, and what does that leave open to build?
It started as my own research for choosing company #3. It turned into something worth sharing.
283 companies funded in 2026, fully filterable. Where venture money went, and where it didn't. Founders read it for the white space. Investors read it for the trend.
Explore the map ->Filter it yourself. Full access is free with the memo.
Writing
One operator memo with real metrics, product decisions, and lessons from 10+ consumer products and 2 exits.
Two Exits Later
Posts sell conclusions. The memo shows the cutting room floor: decision logs, numbers I'm watching, and what I'd steal for the next company.
Subscribe ->Medium
Long-form operator essays on marketplaces, AI in production, and the founder journey. Artifacts built to send to an investor or program partner.
AI coding model costs
Sortable comparison of coding benchmarks and approximate $/task across Codex, Kimi, Grok, Cursor, and DeepSeek.
Open the table →Work with me
Two focused routes for founders who need hands-on delivery or an experienced second set of eyes.
AI automation consulting and implementation
Put one AI workflow or agent system into production, with clear boundaries, reliability controls, and an operating plan.
Explore AI systems ->Think through one hard decision
Focused depth on architecture, product, growth, marketplaces, team setup, or build-versus-buy decisions.
See work routes ->Contact
Founders, funds, and operators working in consumer products, marketplaces, cross-border commerce, or applied AI.
