Writing

Featured LinkedIn Posts

View profile →
Posted on LinkedIn
I spent a fantastic summer at Headline, where I built an internal AI agent for market sizing. 𝗛𝗼𝘄𝗲𝘃𝗲𝗿, 𝗶𝘁 𝘄𝗼𝗿𝗸𝗲𝗱… 𝗿𝗶𝗴𝗵𝘁 𝘂𝗽 𝘂𝗻𝘁𝗶𝗹 𝗶𝘁 𝗯𝗿𝗼𝗸𝗲. I scratched my head more than a few times trying to solve flawed early agent choices and mathematical failures. I figured some of these lessons might be useful to others, so I wrote them up in a blog. In it, I explore both the 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝘀𝗶𝗱𝗲 of building agents and the 𝗶𝗻𝘃𝗲𝘀𝘁𝗼𝗿 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀.
#AIAgents#AgenticAI#VentureCapital
View post →
Posted on LinkedIn
This week, I gave my first big lecture. :) Teaching Data Science at Berkeley, especially when I get to nerd out about Economics while doing it, has made my time here even better. Huge thanks to Eric Van Dusen for the opportunity!
#UCBerkeley#DataScience#Economics

Latest on Medium

View all →

Academic & Professional Writing

Sequential Deep Learning for VC

May 2025
Course ProjectINDENG 242B - Machine Learning and Data Analytics IIUC Berkeley

LSTM/Transformer with fine-tuned BERT embeddings; reported PR-AUC 0.359, precision 0.29; bias and deployment notes.

Deep LearningTransformerBERTVenture Capital

Revisiting Utility Maximization in Incomplete Markets

May 2025
Course ProjectINDENG 222 - Financial Engineering Systems IUC Berkeley

BSDE-based analysis across exponential/power/log utilities with liabilities; simulations and numerics.

FinanceBSDEUtility TheoryIncomplete Markets

Net-Worth Optimization Tool

May 2025
Course ProjectINDENG 243 - Analytics LabUC Berkeley

Multi-asset return modeling + Gurobi optimization (stocks, bonds, real estate, gold, cars); objectives include Sharpe, max return, min risk, utility. Predict than Optimize Pipeline

OptimizationPortfolioGurobiFinance

Decision Transformer: RL via Sequence Modeling — Notes

April 2025
Literature ReviewINDENG 242B - Machine Learning and Data Analytics IIUC Berkeley

Literature summary connecting offline RL and sequence modeling; key takeaways and open questions.

Reinforcement LearningSequence ModelingOffline RL

Market Research: Tools for Data-Driven Investing (PE & VC) (Blackskies)

December 2024
Market ResearchIndustryBlackskies

Landscape, TAM/SAM/SOM sizing, competitive mapping, and compliance considerations for data-driven investing stacks.

Venture CapitalDataAutomationMarket Landscape

Couple Matching via Integer Linear Programming

December 2024
Course ProjectINDENG 240 - Optimization AnalyticsUC Berkeley

Weighted bipartite matching (Gurobi) on speed-dating data; produced 265 optimal matches under constraints.

OptimizationILPGurobiMatching

Predicting Startup Success (≥ $100M valuation)

January 2025
Private & Course ProjectINDENG 242A - Machine Learning and Data AnalyticsUC Berkeley

ML on ~10k VC-backed companies; tabular+NLP features; LR/CART/RF/GBM; impact vs. benchmarks.

Venture CapitalMLNLPClassification

Good, Bad and Complex: Product Design in the Financial Industry

October 2023
Bachelor's ThesisLMU Munich

Bachelor's thesis on how product complexity influences consumer understanding and valuation of financial products (structured products focus).

MicroeconomicsProduct DesignFinanceComplexity

Cycle Time Prediction in Customer-Individual Production Planning Using Machine Learning

May 2023
Bachelor's ThesisTUM

TUM B.Sc. thesis predicting manufacturing cycle times with ML for customer-specific production processes.

Machine LearningManufacturingForecastingProduction Planning

The Market for Quacks (Ran Spiegler, 2006)

Sep 2022
Seminar PaperLMU Munich

Behavioral IO seminar paper on deception/misperception and signaling in markets with "quack" providers.

Behavioral IOInformation EconomicsSignalingDeception

The Liabilities and Risks of State Sponsored Pension Plans (Novy-Marx & Rauh, 2009)

April 2022
Seminar PaperLMU Munich

Seminar paper reviewing underfunding risk and actuarial assumptions in public pensions.

Public FinancePensionsRisk