Air Force Ken

Operating now

Ventures

A CTO seat at NuCalm, and a set of systems I built for decisions that move capital: AI sense-making, natural resource operating pictures, real estate, and trading.

01 / Part-time CTO

NuCalm

NuCalm makes audio you choose by how you want to feel. Put in earbuds, pick a state like Unwind, Focus, or Fall Asleep, and press play. There is nothing to practice. The technology started as a clinical device used in surgery, and since 2021 it has been an app.

I serve as NuCalm's part-time CTO, focused on scaling the business and bringing AI into how it markets. A lot of that starts with being found. More people now ask an AI assistant before they search, so I studied how NuCalm shows up in those answers and built content written for the way 14 different assistants read.

I am also leading a ground-up rebuild of the web experience: a few plain choices, strict rules on what we claim, and copy that sounds like a friend's recommendation. AI does the research, first drafts, and creative. People make the final call.

  • Business scaling
  • AI adoption in marketing
  • AI discoverability across 14 assistants
  • Web experience rebuild
  • Claims discipline
nucalm.com
A listener resting in a leather chair with headphones on and eyes closed
2.7M
Surgical procedures used the technology behind it
36
Years of research
100
Countries
2.5B
Minutes played

Being found by AI

Every assistant, one accurate answer

When someone asks an assistant how to sleep better or calm down before a big moment, the answer should describe NuCalm correctly, in plain words, with only the claims we can stand behind. I mapped how each assistant reads, then wrote for all of them.

AI DISCOVERABILITY14 ASSISTANTS · ONE STORYClaudeChatGPTGeminiGrokPerplexityLlamaMistralDeepSeekQwenKimiMiniMaxCohereNovaNemotronNUCALMONE ACCURATE ANSWER

02 / AI sense-making

300+ AIs, one adjudicated answer

I wired together more than 300 AI models with AI adjudication for sense-making. Each model reads the same evidence on its own. An adjudicator model weighs where they agree, logs where they don't, and scores its confidence. People get a sourced brief and still make the final call.

The same pattern runs nightly research briefs with a memory that carries from one edition to the next, and a review board where independent models each read a setup before a chief adjudicator makes one call.

  • Independent analysts across providers
  • Adjudicator reconciles and scores
  • Dissent logged, not averaged away
  • Memory across editions
  • Human sign-off before anything ships
SENSE-MAKING ENGINE300+ MODELS · ADJUDICATEDSOURCESWEB SEARCHLIVE FEEDSFILINGSRESEARCHMARKET DATAPRIOR BRIEFSADJUDICATEMEMORY · EACH EDITION BUILDS ON THE LASTASSESSMENTAGREEMENTDISSENT LOGGEDEVIDENCE TRAILHUMAN SIGN-OFF

03 / Natural resource acquisition

A common operating picture for the ground truth

I build common operating pictures for natural resource acquisition. Resource sites, concessions, infrastructure, environment, hazards, and security risk sit on one live map, so a team can judge ground conditions before it commits capital.

AI drafts the daily briefing from what the map is seeing. A person approves it before anyone acts on it.

  • 30+ live and reference layers
  • Resource, infrastructure, and environmental overlays
  • Risk, hazard, and geofence alerts
  • AI briefings with human sign-off
  • Export to Google Earth and GIS tools

Illustration uses a fictional region.

COMMON OPERATING PICTURE30+ LAYERS · LIVELAYERSResource sitesConcessionsProtected areasInfrastructureHazardsAviationMaritimeVegetationField devices+ 21 MOREEXPORT KML · GEOJSONPRESENT MODESITE 07 · SELECTEDaccess ok · water okEVENTS 128 · AIRCRAFT 14 · DEVICES 6 · FICTIONAL REGIONFEEDALERTSAISOURCESGEOFENCEDevice left zone 04HAZARDHeavy rain, access roadAVIATIONLow pass near site 11AI BRIEF · DRAFTKEY JUDGMENTSAWAITING HUMAN SIGN-OFF

04 / Real estate investing

Every listing, underwritten overnight

I invest in residential real estate and screen deals with my own tools. A nightly sweep runs the numbers on every listing across more than a hundred markets, maps the risk, and queues outreach only for the deals worth a call.

AI reads the listing remarks. It never sets a number. The math and the final decision stay with me.

  • Nightly multi-market deal sweep
  • Wholesale, flip, and rental math
  • Flood, income, and crime overlays
  • Agent outreach with human review
DEAL DESKUNDERWRITING4 UNITS · 3 LEVELSELEVATION A · ILLUSTRATIVEUNDERWRITINGPURCHASEREHABAFTER-REPAIR VALUERETURN PROFILECAP RATECASH ON CASHDSCRDILIGENCEBuy boxComps pulledRent rollInspectionTitle + liensExit plan

05 / Trading

Commodities, stocks, and futures

I trade stocks, futures, and commodities with software I built: rules-based setup detection, position sizing tied to the live account, and a panel of AI models that checks each call against the method before any order goes in.

Every setup can be replayed against past data, and the AI connector runs paper-only by default with hard caps on order size.

  • Stocks, futures, and commodity contracts
  • Rules-based setup scoring
  • Risk-first position sizing
  • AI review panel with hard guardrails
  • Replay testing against past setups

Nothing here is investment advice.

MULTI-ASSET DESKCOMMODITIES · FUTURES · EQUITIESVOLUMETERM STRUCTUREFRONTDEFERREDCROSS-ASSET WATCHENERGYMETALSGRAINSRATESINDEXEQUITY

Have a problem shaped like one of these?

Operating roles, advisory seats, and builds where the decision matters more than the demo.

Start a conversation