Skip to content

← Back to the map

01 Product Portfolio

Product Portfolio

Most of these products sit inside other systems: claims platforms, health records, factory floors, university back offices, fund pipelines.

A decade-plus of ML/AI products, from health-data pipelines to AI investing tools, including many zero-to-one builds.

Route from 2013 to 2026 with one landmark per chapter of work: Wyndham, Bethune-Cookman, Amida, ServicePower and Monk, Autotech Ventures, and AI tools.

Showing 23 of 23 products

Three inbound streams into one daily triageDeals arrive from Slack, email and meeting notes, merge into one triage step that checks thesis fit and NDAs, split into pass, refer and advance lanes, and end as CRM records.SLACKEMAILMEETING NOTESTRIAGETHESIS FIT · NDA CHECKPASSREFERADVANCECRM RECORDSLOGGED · ENRICHED · MARKED DONE
Investing automationInternal2026SOURCING · DILIGENCE

Deal intake and triage automation, AIC

100+ AI COMPANIES REVIEWED SINCE JULY

Turned partner inbound from Slack, email and meeting notes into a daily triaged pipeline in the CRM, with every deal logged and enriched.

The role

Built at Alpha Intelligence Capital. A Claude-based pipeline pulls inbound deals from Slack, email and Granola meeting notes into Attio and triages every one daily.

More

Each deal gets logged, enriched and scored on thesis fit. NDAs are checked against the firm template, with deviations flagged. Pass and referral notes come back as drafts for review, and processed items get marked so nothing is handled twice.

Key call

Replaced manual CRM entry. Every inbound deal now lands in Attio already logged and enriched.

One query, many sources, one summaryOne query fans out in parallel to web search, social sources and market data. The results come back into one summary with source tags attached.ONE QUERYWEBSOCIALMARKET DATAIN PARALLELSUMMARYWSMSOURCE TAGS KEPT
Investing automationPrototype2026RESEARCH · SOURCING

Startup & Market Research Hub

Puts startup and market research in one search box, with a cited source behind every answer.

Why
By the time a company shows up in a standard database or a funding announcement, the signal is often stale. The hub is built to catch it earlier.
How
A Next.js app. One query runs in parallel across web search, verified social sources and market data. LLMs summarize the results, fall back between models when one fails, and keep the sources attached.
Use
Used for research and sourcing across thesis areas.
Demo
Many sources narrowed to a short ranked listA wide fan of sources passes through a thesis-fit scoring filter. Thousands of signals narrow to a ranked watchlist of qualified deals.935+ SOURCESTHESIS-FIT SCORE2,495 SIGNALSWATCHLIST12345500+ QUALIFIED DEALS
Investing automationInternal2024 to 2026SOURCING

AI sourcing engine, GroundForce Capital

500+ QUALIFIED DEALS · 935+ SOURCES

Generated 500+ qualified deals by scanning 935+ sources and scoring 2,495 company signals on thesis fit.

Read the story

AI products scored on five axes, side by sideSeveral AI products pass through one scoring grid on accuracy, latency, security, integration depth and cost, and come out as a side-by-side comparison for investment committee.AI AAI BAI CSCORING GRIDACCURACYLATENCYSECURITYINTEGRATIONCOSTSIDE BY SIDEAI AAI BAI CLIKE-FOR-LIKE VIEW FOR IC
Investing automationInternal2024 to 2026DILIGENCE

AI diligence toolkit, GroundForce Capital and AIC

70%+ LOWER DILIGENCE COST · 900+ AI COMPANIES

Cut outside diligence costs 70%+ by bringing technical reviews in house, backed by a benchmark that scores AI products side by side.

Read the story

One CIO hub, six portfolio companiesA CIO hub connects to six generic portfolio company tiles, each with a roadmap bar and a tech stack.CO 1CO 2CO 3CO 4CO 5CO 6CIO
Investing automationInternal2024 to 2026PORTFOLIO SUPPORT

Portfolio support toolkit, GroundForce Capital and AIC

15%+ LOWER TECH SPEND · 6 PORTFOLIO COMPANIES

Cut portfolio tech spend 15%+ by auditing roadmaps and tech stacks with executives at 6 portfolio companies, as in-house CIO.

Read the story

A car taken apart, a servicing workflow, and mechanic interviewsAn electric vehicle shown taken apart, a three-step servicing workflow, and interviews with mechanics and technicians.TAKEN APART123SERVICING WORKFLOWMECHANICS · TECHNICIANS
Founding teamArchived2023 to 2024VENTURE STUDIO

RepairWise, Autotech Ventures studio

3 FOUNDING HIRES

Hired 3 founding team members and ran discovery with mechanics and technicians to test remote repair for connected vehicles.

Read the story

Claims split by an ML triage stepA claim passes through an ML triage node. Low-complexity claims route straight to dispatch, others go to human review first. Dispatch then sends a matched technician.CLAIMMLTRIAGELOW COMPLEXITY: AUTO-ROUTEDHUMAN REVIEWDISPATCHSKILL · TRAVELTECHNICIAN
Built and ledShipped2020 to 2024

ML-assisted dispatch and claims automation, ServicePower

18%+ LOWER CLIENT OPEX · 8+ CONTRACTS

Cut client operating costs 18%+ across 8+ contracts with ML-assisted dispatch, built with the CTO and core engineering team.

Read the story

Fourteen firms on one deployment pathFourteen firm nodes, LG, GE and Samsung among them, converge onto a single highlighted deployment path that leads to the Lowe's contract.LGGESAMSUNG14 FIRMSONE DEPLOYMENT PATHLOWE'S
Built and ledShipped2020 to 2024

Fourteen firms, one deployment path, ServicePower

$13M+ ARR · 30% FASTER CLAIMS

Grew ARR $13M+, the company's largest contract, by aligning 14 firms, LG, GE and Samsung among them, on one deployment path.

Read the story

Three Monk Robotics products on one pathA route with three landmarks: a solar mower, a UV sanitizing robot built on the same base after the COVID pivot, and an IoT retrofit kit for factory machines.01 SOLAR MOWERCOVID PIVOTSAME BASE02 UV SANITIZERTO FACTORIES03 IOT RETROFIT
FounderArchived2020 to 2022

Monk Robotics: three products on one base

3 MVPs · 2 ON ONE BASE

Shipped 3 MVPs through COVID: a solar mower, a UV sanitizing robot on the same base, and a factory IoT retrofit.

Read the story

Scheduling from inside the claimAn adjuster in Guidewire ClaimCenter requests a visit. The accelerator passes it to ServicePower scheduling, which books a field adjuster. Appointment status, reschedule and cancel flow back into the claim.CLAIMCENTERADJUSTER + CLAIMACCELERATORREQUEST A VISITSERVICEPOWERSCHEDULINGFIELD ADJUSTERSTATUS · RESCHEDULE · CANCEL
Built and ledShipped2020 to 2021

Field scheduling inside the claims system, ServicePower x Guidewire

3 CLIENT RENEWALS · 3x ADOPTION SPEED

Drove 3 client renewals by taking ServicePower scheduling into Guidewire ClaimCenter, from design to Marketplace launch.

Read the story

From the DC area to Kathmandu and out across NepalFunds raised in DC, Maryland and Virginia move along a route to Kathmandu with 100+ oxygen concentrators, then fan out to local organizations.DC · MD · VA$53,000 RAISEDFREIGHT · 100+ CONCENTRATORSKATHMANDULOCAL ORGANIZATIONS10 L/MIN
VolunteerArchived2020 to 2021

Nepal COVID oxygen logistics

$53K RAISED · 100+ CONCENTRATORS

Raised $53K and delivered 100+ oxygen concentrators to Kathmandu, from fundraising through freight to local distribution.

Read the story

Research and requirements into one format, then a contractResearch, BRD and SRD documents feed one standard proposal format used across 36+ deals, leading to a signed $17M contract.RESEARCHBRDSRDSTANDARD FORMAT36+ DEALS · CMS · DOD · PALANTIR$17M CONTRACT
Built and ledArchived2017 to 2020

Amida proposal engine and the $17M contract

$17M PRIME WON · 30% PROPOSAL EFFICIENCY

Won Amida's largest contract, a $17M federal health-data prime, by chairing the proposal and leading technical scoping and delivery.

Read the story

Point-to-point links replaced by one FHIR layerOn the left, four sources (EHR, Medicare claims, a legacy Oracle database, a payer model) wired to three apps make twelve links. On the right, the same sources and apps connect once each through a FHIR layer.POINT TO POINTEHRMEDICAREORACLEPAYER12 LINKS TO BUILDONE FHIR LAYEREHRMEDICAREORACLEPAYERFHIRVIEWERCAREANALYTICSONE CONNECTION PER SYSTEM
OwnedArchived2017 to 2020

FHIR/EHR interoperability product line, Amida

100%+ YoY GROWTH · $17M PRIME

Owned the FHIR product line that moved Medicare claims, military health records and EHR data between systems for CMS, MHS and UPMC programs.

Read the story

Golden and infected chips compared by an ML modelTwo FPGA boards side by side, a clean golden model and an infected copy. Instrument probes feed an ML model that reports detection, location, impact and a Trojan signature.GOLDEN MODELINFECTEDPROBESMLREPORTDETECTLOCATEIMPACTSIGNATURELEAKER · MODIFIER · RELIABILITY
OwnedPrototype2017 to 2020

Finding hardware Trojans in chips with ML

NAVY SBIR PHASE I · 10 TROJANS MODELED

Owned an ML product line that spots hardware Trojans in chips by comparing a clean emulated system against infected ones, funded through Navy-supported SBIR work.

Read the story

VA security IV&V: reviews into a handbookVA security policy, the SDLC, GRC tools and the acquisition process feed one IV&V review, which produces a security implementation handbook.POLICYSDLCGRC TOOLSACQUISITIONIV&V REVIEWSECURITY HANDBOOKFINDINGS INTO ENGINEERING FIXES
Built and ledArchived2017 to 2020

VA security IV&V, Amida

Led independent verification and validation of VA's security process, from policy and SDLC review to a security handbook teams could follow.

The role

Reviewed VA's security laws and policies, the SDLC, GRC tools and the software acquisition process. Designed report templates and a security implementation handbook, defined GRC roles and tracked completion of security functions.

Key call

Turned findings into engineering fixes. Each IV&V finding became a fix the teams could act on.

The Data Reconciliation Engine in the middleHealth records from many sources pass through one adapter layer. The highlighted middle layer is the Data Reconciliation Engine, which reconciles them before they reach other systems.RECONCILE (DRE)MANY SOURCESSYSTEMS + APPS
Took to marketOpen source2017 to 2020

Data Reconciliation Engine, Amida

Took an open-source engine to market that reconciles one patient's health records across systems: requirements, user research, demos.

The role

Wrote the requirements, ran user research, wrote technical volumes for proposals that extended it, and demoed it to potential partners and clients.

Key call

Served as the liaison between engineering and the business side. Turned what users and buyers asked for into functional and technical requirements.

GitHub
Orange at the end of the health data flowHealth data passes through one adapter layer to the systems and apps that use it. The highlighted app at the end is Orange, for logging medication.ADAPTER LAYERMANY SOURCESORANGE APP
Took to marketOpen source2017 to 2020

Orange, Amida

Took an open-source medication adherence app to market, writing the requirements and running user research and demos.

The role

Wrote the requirements, ran user research, wrote technical volumes for proposals that extended it, and demoed it to potential partners and clients.

Key call

Ran the feedback loop. Took user research back to the engineering team and kept it moving.

GitHub
INDABA at the start of the health data flowData from many sources passes through one adapter layer. The highlighted start of the flow is INDABA, where data gets collected.ADAPTER LAYERCOLLECT (INDABA)SYSTEMS + APPS
Took to marketOpen source2017 to 2020

INDABA, Amida

Took an open-source data collection and project management tool to market, from requirements through client demos.

The role

Wrote the requirements, ran user research, wrote technical volumes for proposals that extended it, and demoed it to potential partners and clients.

Key call

Wrote proposals that extended the product. Each technical volume scoped new features on top of the existing codebase.

GitHub
From paper to one digital flowBefore: a stack of printed forms and signatures. After: one digital flow from form to approval to purchase order, with processing time measured at each step.BEFOREPRINTED FORMSPAPER APPROVALSAFTERFORMAPPROVALPOPROCESSING TIME MEASURED AT EACH STEP
Built and ledInternal2014 to 2017

Getting a university off paper, Bethune-Cookman

$6.9M+ SAVED · 33% EFFICIENCY

Saved $6.9M+ and lifted operational efficiency 33% by building the university's first data and digitalization function, reporting to the CFO.

Read the story

Search queries to segments to likely buyersSearch queries flow into customer segments by income, age, premium and location. Regression models rank the segments by likelihood to buy, and the top one gets the SEO focus.SEARCH QUERIESSEGMENTSINCOMEAGEPREMIUMLOCATIONREGRESSION MODELSLIKELY TO BUYFOCUS SEO HERE
FellowArchived2016

Marketing analytics and SEO, Policygenius

Showed which searchers were most likely to buy insurance, so SEO content could target them. Northwestern fellowship.

The role

Policygenius sold insurance online. The work ran across its marketing and customer datasets in SAS.

More

Segmented customers by household income, age, premium and location, then built linear and logistic regression models to see which segments converted. Those segments mapped back to the search terms and landing content that brought them in, so the marketing team could focus SEO content on the visitors most likely to buy.

Key call

Tied the segments back to search. That told the marketing team which SEO content to focus on.

Social posts turned into a needs mapScattered social media posts about rescue and relief needs flow through a Hadoop pipeline (Pig, Hive, HBase, Sqoop) into a map with clustered pins that NGOs used to prioritize.SOCIAL POSTSHADOOPPIG · HIVEHBASE · SQOOPNEEDS MAPNGOS PRIORITIZE
VolunteerArchived2015

Nepal earthquake data mapping

$45K+ RAISED

Mapped social media rescue requests so NGOs could prioritize relief after the 2015 earthquake, and raised $45K+ plus supplies.

Read the story

A crowded timetable becomes a balanced oneOn the left, the Fall 2014 class schedule crowds into a few slots. An R and SQL model produces the schedule on the right, spread evenly, with a 68% efficiency gain.FALL 2014 BASELINER + SQL MODELMODELED SCHEDULE68% EFFICIENCY GAINFOUR YEARS OF SCHEDULING DATA
Built and ledInternal2014 to 2015

B-CU scheduling optimization

68% EFFICIENCY GAIN

Gained 68% scheduling efficiency over the Fall 2014 baseline by modeling four years of class data in R and SQL.

The role

Pulled the scheduling data from the university's MySQL server with SQL, then wrangled and modeled it in R.

A mapping tool handed to a pilot teamA tablet showing an interactive map, with lines out to three people on the pilot team who were trained on it.PORTABLE MAPPILOT TEAMTRAINED ON THE NEW TOOL
InternArchived2013

Wyndham interactive mapping pilot

Trained the pilot team on a portable interactive mapping tool, then built the timeshare business case for Nepal and sent leads to RCI India.

The role

Summer internship at Wyndham Worldwide / RCI. Implemented a portable interactive mapping technology and trained the pilot team. Then built a timeshare feasibility study and business model for Nepalese markets and forwarded leads to the RCI India team.

Earlier: graduate data science work in R and SAS, on GitHub.