MATTHEW ADNER

Summer Analyst

TVG Consulting – Remote

• Built an AI tool to collect, clean, and integrate public data into a proprietary dataset for market research.
• Contributed to buy- and sell-side due diligence in digital infrastructure, focusing on growth, valuation, and synergy.
• Informed strategic and operational recommendations through market research and expert interviews; supported client relations.
• Developed financial models and performed geospatial and data-driven analyses to support M&A strategy and execution.

Co-President; Quantitative Strategies Group Founder/Leader

Student Investment Fund at Union College – Schenectady, NY

• Lead meetings of 50+ members, set agendas, coordinate investment discussions, and ensure productive collaboration among sector leaders and their analyst teams.
• Co-manage over $1,000,000 of the college’s endowment in an equity/fixed income portfolio.
• Founded group within the fund dedicated to investing based on quantitative strategies rather than fundamental analysis. View slides.
• Practiced backtesting, trading analysis, and portfolio optimization, improving data manipulation, analysis, and visualization skills.
• Explained technical processes and systems to non-technical audiences in an approachable manner.
• Formerly Real Estate sector leader, where my team and I pitched a REIT. Read about it here.
• Presented on financial derivatives, not including options. View slides.
• Presented on options, focusing on basics and teaching a rough price intuition. View slides.

Research Assistant to Prof. Brian Melzer

Tuck School of Business at Dartmouth College – Hanover, NH

• Analyze large sets of real estate price data to identify trends and patterns.
• Translate successful equity market trading strategies to real estate markets and backtest their effectiveness.
• Develop R code to synthesize data and findings into clear visualizations and dashboards.
• Synthesized real estate transaction data into MBA case study materials on investment strategy and property valuation.

Assistant Scientist and Mentor

NH Academy of Science Summer Program – Lyme, NH

• Taught modules on computer science, machine learning, computer vision, 3D modeling, and Arduino for students aged 13-18.
• Trained students in research and lab procedures, advising experiment design, execution, and analysis.
• Co-managed social media accounts, web presence, and directed various outreach efforts nhacadsci.org.

Teaching Assistant: Introduction to Computer Science

Hanover High School – Hanover, NH

• Designed exercises, presented lectures, and worked with students individually and in small groups.
• Graded assignments and developed software for student interaction measurement.
• Developed software for student interaction measurement. Project GitHub Repository.

Ski Instructor

Dartmouth Skiway Snowsports School – Lyme, NH

• Led beginner and intermediate ski lessons for individuals and groups of all ages.
• PSIA Alpine Ski Level 1 Certified.

Student Researcher

New Hampshire Academy of Science – Lyme, NH

• Delegate to AAAS affiliated AJAS conference for 3 consecutive years.
• Created research abstracts, presented in poster sessions and roundtables.
• Read abstracts: 2019, 2020, 2021.

Quantitative Trading Algorithm and Backtesting System

This project is a custom-coded algorithmic trading system I built from the ground up using python and the libraries Pandas, Numpy, Plotly. The goal was to create a tool that could scan the market (over 16 million datapoints), identify promising opportunities based on indicators, and track simulated trades over time.

Data Processing
This component cleans and aligns data from multiple sources while handling messy edge cases, like companies with multiple share classes or that didn't have historical price data. The data is formatted as a 3-dimensional dataframe with axes representing different stock tickers, different attributes like price or indicators, and time. This structure allows operations to be far more computationally efficient.

Custom Indicators
One of the main signals I built measures how far a stock is trading outside of its usual range. It combines rolling linear regression and standard deviation calculations to flag when price action deviates significantly from recent trends.

Stock Selection Logic
The strategy filters the entire market down to a refined list of tradable stocks by combining a custom signal designed to detect abnormal price behavior, other quantitative indicators, and a set of rule-based screens. These filters help exclude companies with undesirable attributes such as low liquidity or companies headquartered in tax haven countries, and apply rolling calculations to smooth out noise. The result is a focused subset of equities that stand out for their risk/reward profile.

Portfolio and Trade Tracking
The portfolio system consists of several interrelated classes. At its core are classes that manage individual positions and full portfolios, allowing for accurate tracking of trades, position-level metrics, and portfolio-level performance over time. These classes are designed to simulate realistic trading behavior, including variable position sizing, dynamic position entries and exits based on current unrealized profits, losses, and changes in the indicators that informed the position entry. The simulation also accounts for variable trading costs based on company size and trading volume, allowing for a more accurate reflection of real-world execution and slippage.
In addition, a trade analysis class tracks every trade made during the simulation. Once the backtest is complete, it performs a set of analyses on portfolio outcomes and trade patterns. It allows users to view performance based on a variety of attributes including sector, market cap, indicator, and trade duration, making it easier to see where profits and losses came from and which subsets of the market the strategy was most profitable over.
Together, these components make it easy to understand how the strategy behaves across different market conditions and help refine future iterations.

Next Steps
While building the backtest system was a great learning experience, future versions will use prebuilt frameworks to focus more on strategy development. One planned feature is applying NLP to SEC filings. By training a language model on company reports and measuring shifts in perplexity, I hope to detect changes in tone that signal upcoming stock reversals.

Real Estate Momentum Dashboard

This project is an interactive dashboard that I built in R with the ggplot2, tidyverse, DT, and shiny libraries to help visualize momentum trends in U.S. real estate markets for Professor Brian Melzer (Tuck School of Business, Dartmouth College). It identifies which areas are likely to continue outperforming based on past performance and other variables. The dashboard allows users to choose features to train a linear model to predict the performance of real estate markets over the next term (also selected by the user). This data is visualized as a heatmap of predicted top-performing U.S. markets over the next term.

Data Sourcing and Processing
This tool leverages Zillow data on average home prices and rents in different real estate markets, combined with CPI data from the Bureau of Labor Statistics to adjust prices for inflation. Each real estate market in the dataset corresponds to a specific city or county in the U.S. Historical housing and rent data were cleaned and aligned across hundreds of markets. From the raw price data, forward returns (used as prediction targets) and past returns (used as features) were computed for various time horizons. To allow for inflation-adjusted comparisons, prices were converted to real terms using CPI multipliers tied to the selected date.

Predictive Features and Targets
The dashboard supports flexible feature selection, allowing the user to pick which of the following features they want the model to use for prediction:

• Past returns (6-month, 1-year, 5-year)
• Percentile of Past returns (6-month, 1-year, 5-year)
• Price-to-rent ratio
• Real or nominal price
• Region size rank

The user can pick which of the following targets they want to predict:

• Future returns (6-month, 1-year, 5-year)
• Percentile of Future returns (6-month, 1-year, 5-year)

These features and targets are passed into a linear model trained on historical data available before the selected date, which then outputs a ranked list of expected returns by region.

Visualization
The primary output of the dashboard is a heatmap of the U.S. that shades states by their predicted average return over the forward-looking prediction term (6-month, 1-year, 5-year). Below the map is a sortable table that displays the top regions along with key attributes including those used in the prediction.

Insights and Applications
This dashboard makes it easy to spot where momentum is strongest across U.S. housing markets. Users can fine-tune predictions by layering in valuation metrics like price-to-rent ratio and real price levels, helping surface regions that may be both trending and undervalued. The tool provides a fast, intuitive interface for exploring these patterns and testing custom forecasting setups with just a few clicks.

Consumer Review Bias Sentiment Analysis

Hanover, NH

Used sentiment analysis and machine learning to identify statistically significant consumer bias given to smaller companies on star ratings of Amazon reviews.

Read the paper

Project GitHub Repository

Bachelor of Science in Mathematics and Economics with a Minor in Computer Science

Union College – Schenectady, NY

• Awarded Presidential Scholarship for merit.
• Yoga Club Treasurer – Managed budget, coordinated logistics for instructor contracts, and transportation for on and off-campus sessions (Jul 2025 – Jun 2026).
• Beuth House Procurement Chair – Identified and procured resources for events, negotiated contracts with local businesses, collaborated with faculty (Mar 2023 – Nov 2024).

Summer@INSEAD

INSEAD Business School – Fontainebleau, France

Two-week intensive program focused on business strategy, management, and operations.

High School Diploma

Hanover High School – Hanover, NH

• On the swim team for 4 years, elected team captain in senior year.
• Computer Science Teaching Assistant in senior year.
• Founded Hanover High School House Plant Club and grew membership to over 50.