Franklin at Little Island, the Lower Manhattan skyline behind him

Franklin “Trip” Dickinson

Founder · Monomoy Studios
Cornell Tech · M.Eng. CS '26 · 3.8 GPA
New York, NY
Meta Platforms, Civis Analytics
Dartmouth College · BA
US Citizen

Data scientist who builds. Seven years in political and media data, forecasting elections at Civis Analytics and measuring campaigns at Meta.

At Civis, I built election models from the national voter file and survey data and forecast 6,000+ state legislative races in 2020 for the DLCC. At Meta, I measured national messaging campaigns with difference-in-differences, built A/B testing from scratch for a research app, and co-built a media intelligence dataset five teams adopted as their standard.

Now, after an M.Eng. in Computer Science at Cornell Tech, I build and evaluate LLM products. The main one is Ask Fogo, an AI money companion in beta.

Monomoy Studios

Founder · June 2026 – present · New York

Monomoy Studios is my independent AI product studio. I do the product thinking, design, engineering, and testing myself, with AI coding agents (Claude Code and Codex) working in parallel so one person can ship what used to take a team.

The main product is Ask Fogo, in beta. Below it are product prototypes from the studio. The projects further down are from my Cornell Tech M.Eng. in Computer Science.

Ask FogoBeta
AI money companion

Money talks. Fogo listens.

Ask Fogo connects to your accounts, gets the transactions right, and tells you "you can afford this" or "I don't know." I built the first version during my M.Eng., then worked on it full-time for three months after graduating.

What it does
  • Connects your accounts through Plaid and sorts every transaction, with no tagging
  • Tracks net worth over time and separates your own saving from market moves
  • Ask Fogo: a chat backed by 30 tools that only states numbers it can trace, and says "I don't know" when it can't
  • A month-end review that checks in on the plan you actually accepted
  • Household mode, and life events like a raise or a move built into the plan
How it's built
  • Solo, with Claude Code and Codex working in parallel: 3,400+ commits and 1,160+ merged pull requests in three months
  • Classification runs rules, then a cache, then an LLM for only the hard cases
  • SQL computes every number; the model only explains it, and a validator blocks any figure it can't trace
  • LLM-judge evals calibrated against my own blind ratings, 12,000+ automated tests, and Ember, an agent that tests the live app on a schedule
Product prototypes from the studio

A real estate platform for Tulum, Mexico, a market with no MLS and few street addresses. Cotejo collects every long-term asking rent from listing portals and Facebook groups, matches each one to its building and neighbourhood, and shows the count behind every figure. Open Cotejo ↗

Cotejo Tulum: median monthly asking rents by neighbourhood on a map and bar chart

Built at Cornell Tech

M.Eng. in Computer Science · Certificate in AI for Engineering · 2025–2026

Experience & education

Jun 2026 – Present

Monomoy Studios

Founder · one-person product studio
New York, NY · askfogo.com

Building Ask Fogo full-time: grounded AI answers, a decision layer, and agentic QA, shipped solo with parallel AI coding agents.

Product studio, full-time since graduation. Main product: Ask Fogo, an AI money companion in beta (FastAPI, React, PostgreSQL, Claude, Langfuse, Plaid).

  • –Shipped 3,400+ commits and 1,160+ merged pull requests in three months, running parallel AI coding-agent lanes as a solo developer
  • –Built the grounding layer: every figure carries where it came from, and a reply validator makes sure Fogo never states a number it can't back, with "unknown" as an allowed answer
  • –Built the decision layer: recommendation contracts and briefs, decision memory, and a belief ledger with an outcome loop that scores whether Fogo was right
  • –Stood up agentic QA: Ember (nightly Playwright crawl, Claude-vision review, GitHub issues, auto-fix lane), a persona dogfood runner, a ten-journey launch gate, and FogoBench
Grounded AI Agentic QA Product design Solo production
Aug 2025 – May 2026

Cornell Tech

M.Eng. Computer Science · 3.8 GPA · graduated May 2026
New York, NY · Certificate in AI for Engineering

The modern AI stack, a Certificate in AI for Engineering, the first version of Fogo, and a first-place hackathon win.

M.Eng. focused on the modern AI stack: generative modeling, ML engineering, and trustworthy AI.

  • –Solo-built the first version of Fogo, a production AI personal finance companion, from design through deployment
  • –Won 1st place at Empire Hacks, an AI hackathon, with Glassbox: AI transparency tooling for medical diagnostics, out of 40+ teams
Generative modeling Trustworthy AI ML Engineering 1st place, Empire Hacks
Mar 2022 – May 2025

Meta Platforms

Data Scientist, Product Analytics
New York, NY

Campaign measurement for Public Affairs, then experimentation and analytics for Meta’s survey products.

Did the campaigns work, who are the advocates, and what is the news saying about Meta.

Embedded data scientist for Public Affairs, working with the Public Affairs Marketing and Survey and Demography Science teams on causal measurement, marketing analytics, and media data products.

  • –Quantified 18–28pp brand health lift from messaging campaigns using regression adjustment and diff-in-diff
  • –Designed a 7-cohort experiment to identify brand advocates on-platform, surfacing a segment with 5x the response rate of the general population
  • –Helped build Headline Metrics, an ML text classification system for tracking news on and off platform: co-specced and prototyped it, PM'd the build, and readied it for production before handing it to data engineering. It replaced the org's manual topic tracking and became the company standard for 5+ teams

A/B testing built from scratch, regressions root-caused, and impact analyses for platform changes.

Embedded data scientist for the Survey Platform and Viewpoints teams: ran experiments, investigated falling survey start rates as liaison to the Survey and Demography Science researchers, and handled SEVs. Owned the data layer, partnering with PM, data engineers, and software engineers.

  • –Built the experimentation infrastructure from scratch, defined metrics, and ran A/B tests on the onboarding funnel, driving +2.9% first login
  • –Root-caused multiple verification and onboarding regressions by segmenting cohorts across the user funnel and correlating against deployment timelines; escalated reproduction reports that unblocked engineering
  • –Owned impact analyses for product and change-management decisions, quantifying how platform changes would affect the survey data researchers depend on
Experimentation at scale Causal inference ML platform development Root cause analysis
Aug 2018 – Jan 2022

Civis Analytics

Senior Applied Data Scientist, Politics
Washington, D.C.

Forecasting state legislative races and testing political ads at national scale, delivered straight to clients.

Political data science consultancy. Applied research work: forecasting state legislative races, running advertising experiments at national scale, and delivering findings directly to clients.

  • –Led district-level forecasting for state legislative races: translated opinion models into client-ready insights and delivered findings across election cycles, including building the data pipelines
  • –Ran A/B tests on national-scale political advertising programs. Prediction errors had direct electoral consequences.
Election forecasting Political ad testing Client delivery Client management
Oct 2016 – Mar 2018

QuanticMind

Revenue Intelligence Analyst
Buenos Aires, Argentina

My first job: CRM data, lead scoring, and sales forecasting for every revenue team.

Embedded analytics at a paid search ML startup (SEM).

  • –Owned CRM data infrastructure and built lead scoring models from LinkedIn data
  • –Managed sales forecasting covering every revenue function: sales, marketing, SDRs, and AEs
Lead scoring Sales forecasting CRM analytics Paid search
Sep 2012 – Jun 2016

Dartmouth College

BA, Government & Anthropology
Modified with Economics · Hanover, NH

Government and anthropology, modified with economics, with study at LSE and the University of Auckland.

Cross-disciplinary foundation built across Government, Anthropology, and Economics.

  • –Grounding in political systems, cultural analysis, and economic reasoning through both quantitative and qualitative research methods
  • –Study abroad at the London School of Economics and University of Auckland
Political economy LSE exchange Quantitative methods Cross-disciplinary

Certificate in AI for Engineering

A Cornell Tech program certificate, earned alongside my M.Eng. in Computer Science · 2025–2026

◆ AI / ML systems
Machine Learning EngineeringMiniTorch · CUDA kernels
+
Built end-to-end ML pipelines covering data ingestion, training, model deployment, and production monitoring. Left with a practical framework for shipping models that hold up outside the Jupyter notebook.
MiniTorchAutogradCUDANumbaGPU optimizationModel deployment
Generative ModelingDiffusion · VAE · GAN · RLHF
+
Implemented diffusion models, VAEs, and GANs from mathematical foundations up, gaining hands-on fluency with the architectures behind modern AI image and text generation. Goes beyond using these systems to understanding how they actually work.
VAEGANDiffusionFlow matchingRLHFSFTDPOKTOMistral 7B
Language ModelingTransformers · tokenization · LLM eval
+
Built and evaluated the core systems behind AI assistants, coding agents, and search: from tokenization and model design through task formulation and real-world evaluation. Gives me the depth to reason about LLM architecture decisions, not just prompt-engineer on top of them.
LLMsTokenizationTask formulationEvaluationArchitectures
Deep LearningCNNs · transformers
+
Implemented CNNs, RNNs, and transformer architectures from scratch in PyTorch, training neural networks on real datasets across vision and language tasks. The foundational course that made every other part of the AI stack legible.
PyTorchCNNsRNNsTransformersComputer visionNLP
Computer Systems & ArchitectureCPU architecture · ML inference
+
Studied how software runs on hardware at the processor level: memory hierarchies, instruction pipelines, and performance optimization. Directly informs how I reason about latency, throughput, and cost when designing ML inference systems.
Memory hierarchiesInstruction pipelinesLatencyThroughputPerformance
◆ AI safety & responsibility
Trustworthy AIGCG jailbreak · PGD attacks
+
Built and attacked ML models against adversarial inputs, privacy exploits, and fairness failures, learning both how to construct robust systems and exactly how they break. Shmatikov's security-researcher lens on AI now shapes every product risk call I make.
FGSMPGDGCG jailbreakAdversarial robustnessMachine unlearningRL alignment
AI Law & PolicyEU AI Act · AntiManipulate.me
+
Analyzed the legal and regulatory landscape governing AI deployment: from the EU AI Act to liability doctrine to data privacy frameworks. Lets me have real conversations with legal and compliance teams about what AI products can actually ship.
EU AI ActAI liabilityData privacyManipulation taxonomyRegulatory compliance
Ethical Decision MakingNormative reasoning
+
Applied normative ethics frameworks to real business cases, building structured approaches to the "should we build this" questions that come up constantly in AI product work. Moves ethical reasoning from gut instinct to defensible argument.
Ethics frameworksCase analysisProduct ethics
Product & entrepreneurship
Product StudioML for medical diagnostics
+
Led the ML modeling track on a cross-functional team with medical researchers, building a machine learning system for novel blood coagulation testing and validating the approach with physicians. Real product work: clinical use case identification, stakeholder alignment, and a path to market.
ML modelingMedical AICross-functionalStakeholder management
Startup StudioFirst CMO · brand identity agent
+
Building an AI product from scratch with a cross-disciplinary team, running customer discovery interviews and opportunity sizing before writing a line of code.Worked on Project Zeta, exploring AI solutions in digital marketing.
Customer discoveryOpportunity sizingAI productGo-to-market
Early AdoptersGTM · customer acquisition
+
Developed and tested real customer acquisition strategies for early-stage products, including actual outreach, contract structures, and system designs for landing first customers. Operationalizes the gap between "people say they want this" and "people will pay for this."
Customer acquisitionOutreachContract designEarly-stage growth