Portfolio

Work That Delivers

AI solutions that produce measurable business outcomes — not just technology demos.

Pagenta
AI Infrastructure / SaaS

Pagenta

Shipping a small app still means hosting setup, a deploy pipeline, and DevOps glue — the gap between "the agent generated this" and "people can actually use it" kills momentum.

HTML to live URL in secondsMCP-native: Claude, ChatGPT, CursorSecrets, cron, push, and state built in
  • One MCP round-trip deploys a real app to a real URL — no editor, no pipeline
  • Persistent state, scheduled cron jobs, and push notifications without extra infrastructure
  • Password-protected pages and shared portals for client and team delivery
  • Built-in analytics on every deployed page, plus CLI and dashboard access

Solution

An agent-native delivery layer that turns a single URL or description into a live, stateful application — letting AI agents and humans ship and manage real apps in seconds

Try Demo
Plurality
AI / Decision Intelligence

Plurality

A single model answers in confident prose whether it knows or is guessing. For high-stakes calls, leaders need to see where the models agree, where they split, and why — not one opaque verdict.

30 frontier models across 11 vendors7 structured debate modesReconciled answer in under 10s
  • Sends one prompt to many models in parallel to avoid anchoring bias
  • Seven deliberation modes — Consensus, Debate, Forecast, Red Team, Oracle, Peer Review, Quick Take
  • Reconciliation report maps agreements, dissents, and which model cited which evidence
  • Custom or preset panels drawn from Anthropic, OpenAI, Google, Meta, xAI, and more

Solution

A reasoning layer over 30 frontier models that routes a question to a panel and returns one reconciled, auditable answer — turning model disagreement into a usable confidence signal

Try Demo
Investment Analysis Engine
Finance / FinTech

Investment Analysis Engine

Institutional-grade equity research requires synthesizing quantitative fundamentals, qualitative intelligence, and macro context — a process that takes analyst teams days per ticker.

Hybrid quant + qual analysisMulti-agent orchestrationSEC XBRL + live market data
  • Deterministic committee scoring (70% weight) ensures reproducible quantitative verdicts
  • LLM synthesis (30% weight) adds qualitative research from earnings calls, analyst reports, and competitive dynamics
  • Monte Carlo simulation with macro regime classification for probability-weighted returns
  • Evidence validation with SHA256 hashes and SEC XBRL cross-checks

Solution

A multi-agent system on Microsoft Agent Framework that generates institutional-grade investment reports combining DCF valuation, scenario modeling, options analysis, and AI-powered qualitative research

Mobius Carbon
Real Estate / Energy

Mobius Carbon

Institutional real estate owners needed finance-grade energy data infrastructure — not another scenario tool, but settlement-aligned clearing and reconciliation across 200+ properties.

6 wks → 2 days35M+ data pointsDiligence-ready output
  • Verified outcomes anchored to authoritative settlement data, not estimates
  • Ranked actions prioritized by financial impact on Net Operating Income
  • Clear accountability — every action has an assigned owner and due date
  • Diligence-ready evidence supporting lender and investor reporting

Solution

The energy operating layer for institutional real estate — trusted baselines, ranked actions tied to NOI, and audit-ready evidence for lender and investor reporting

360 Intelligence
Marketing / SaaS

360 Intelligence

Marketing teams juggle separate tools for surveys, site analytics, SEO monitoring, GEO tracking, visitor recording, and heatmaps — fragmented data, fragmented decisions.

All-in-one marketing intelligenceAI-powered survey generationSEO + GEO monitoring
  • Survey engine with 35+ question types, conditional logic, branching, and AI generation
  • Visitor recording and heatmap tracking for UX insights
  • SEO and GEO monitoring with automated alerting
  • Real-time analytics dashboard with multi-channel distribution

Solution

A unified marketing intelligence platform combining surveys, site analytics, visitor recording, heatmaps, and search monitoring in a single stack

Iceland Economic Signals
Finance / Macroeconomics

Iceland Economic Signals

Iceland’s macroeconomic data is scattered across 235+ Statistics Iceland datasets with no real-time signal layer. Analysts, traders, and policymakers lacked computed indicators tuned to small open economy dynamics.

19 economic signals87.5% policy accuracyReal-time MCP delivery
  • 19 professionally-calculated indicators across inflation, labor, trade, housing, and policy
  • 87.5% validated accuracy on Central Bank rate decision predictions (Oct 2024–Oct 2025)
  • MCP integration gives Claude direct access to real-time Iceland economic conditions
  • REST API, Streamlit dashboard, and CSV/PDF export for institutional delivery

Solution

A signal-based macroeconomic intelligence platform built on FastAPI + FastMCP, transforming raw Statistics Iceland data into actionable indicators with ML-assisted policy probability scoring

Olympus Stack
Developer Infrastructure

Olympus Stack

AI-powered development workflows are fragmented across disconnected tools, and connecting many MCP servers creates unsustainable token overhead — 100+ tools consume ~17K tokens before any work begins.

88% token reduction13 specialized MCPs200+ tools unified
  • Gateway pattern routes to 13 MCPs on-demand, cutting token overhead from 17K to 2K
  • Real infrastructure testing (Colosseum) with Docker-based services — no mocks
  • Multi-model consensus (Delphi) reduces hallucination through cross-model validation
  • Integrated security scanning, code generation, knowledge graphs, and observability

Solution

A modular framework of 13 specialized MCP servers orchestrated through a token-efficient gateway — combining security scanning, multi-model consensus, real integration testing, and AI-powered code generation into one ecosystem

LayerDoc
Productivity / SaaS

LayerDoc

Collaborative writing tools either lack AI assistance or force cloud-only workflows. Teams need rich editing with AI help that works both in-browser and as a native desktop app with local-first storage.

AI writing assistanceMCP integrationDesktop + web app
  • Rich text editing with TipTap and real-time collaboration
  • AI writing assistance powered by MCP integration
  • Electron desktop app with local-first storage
  • Works in-browser or as a native application

Solution

A collaborative writing platform with AI assistance and MCP integration — built as both a Next.js web app and an Electron desktop app with local-first architecture

AI Without the Hype
Media / Publishing

AI Without the Hype

AI coverage is dominated by sensationalism and engagement-driven narratives. Builders and business leaders needed a calm, credible source — one that could operate autonomously at daily publishing cadence.

Fully autonomous daily publishingSelf-governing editorial boardHype scoring on every article
  • Multi-agent editorial board debates coverage and evolves editorial policy
  • Daily briefings produced end-to-end without manual intervention
  • Hype scoring system (0–10) grades every claim across five dimensions
  • Transparent attribution: Produced, Orchestrated, or Written

Solution

A self-governing AI newsroom with autonomous editorial evolution, RSS monitoring, multi-persona essays, and radical transparency about what’s automated

Try Demo

Plus more across hospitality, analytics, autonomous research, and consumer apps — 20+ shipped and counting.

Strategy + delivery.

30 minutes.
You'll know in the first five.

Tell me what you're trying to ship. I'll tell you whether I'm the right fit — and if not, who is.

No pitch. No slides. Just clarity.