Where Projects Fit in the Hiring Picture
To be direct about what matters most in quant recruiting: school name, degree, GPA, internships, and competitive achievements (math olympiads, Putnam, Kaggle) carry far more weight than personal projects. These are the primary signals firms use to decide who gets an interview. A published research paper in a reputable journal is also a strong signal, since it has already been independently validated.
Projects sit below all of these. In the era of LLMs, the actual contribution behind a project is difficult and costly for reviewers to verify, which makes them a weaker signal than credentials that have been externally validated. That said, projects are not worthless. They can signal genuine interest, serve as useful conversation topics during interviews, and help an otherwise borderline candidate secure an online assessment. For candidates without top credentials or prior internships, they are one of the few ways to demonstrate capability. The key is to be realistic about what they can and cannot do for your application.
What Makes a Project Worth Including
If you do include projects on your resume, most will be ignored: another Black-Scholes implementation, another basic moving-average backtest. The ones that might actually come up in conversation share a few traits:
- They solve a non-trivial problem end-to-end, not just one function
- They show awareness of real-world issues (slippage, data quality, survivorship bias)
- The code is clean, tested, and well-documented
- Results are honestly presented with limitations and failure modes
- The write-up reads like a research note, not a tutorial
Research Idea 1: Measuring Backtest Overfitting
Replicate and extend the work of Marcos Lopez de Prado on the probability of backtest overfitting. Apply combinatorial purged cross-validation to a set of trading strategies across equity and futures markets. Quantify how Sharpe ratios degrade out-of-sample as a function of the number of strategy variations tested. Present results with rigorous statistical methodology.
This is publishable research if done well. It addresses a real problem the industry cares about and demonstrates statistical rigor, programming ability, and intellectual honesty. A paper submitted to the Journal of Financial Data Science or similar venues carries far more weight than a GitHub repo.
Research Idea 2: Empirical Analysis of Market Microstructure
Use publicly available LOBSTER data or crypto exchange feeds to study how order book dynamics predict short-term price movements. Measure the predictive power of order flow imbalance, queue position, and spread dynamics across different market regimes. Compare your findings against published microstructure literature (Cont, Stoikov, Cartea).
This type of empirical research is directly relevant to market-making and HFT desks. A well-written paper with clean methodology submitted to Quantitative Finance or Market Microstructure and Liquidity demonstrates you can do the actual work these firms hire for.
Research Idea 3: Price Impact Models and Execution Cost Analysis
Study how trade size, timing, and market conditions affect price impact. Replicate the Almgren-Chriss framework, then test it empirically against real execution data (available from WRDS or crypto markets). Measure how well square-root-law and transient impact models hold across asset classes and volatility regimes.
Execution research is underserved in academia but highly valued by the industry. A paper comparing theoretical impact models to empirical evidence, submitted to a journal like Mathematical Finance or the Journal of Trading, signals deep understanding of a problem every systematic fund faces daily.
Research Idea 4: Factor Crowding and Return Predictability Decay
Investigate how the publication and adoption of well-known factors (value, momentum, quality) has affected their out-of-sample returns. Use CRSP/Compustat data to measure factor Sharpe ratios in pre-publication vs post-publication periods. Test whether crowding metrics (short interest concentration, ETF flows, factor correlation clustering) predict periods of poor factor performance.
This is a live research question at firms like AQR, Two Sigma, and Dimensional. A rigorous paper on factor decay with proper statistical controls is publishable in the Journal of Financial Economics or Review of Financial Studies and demonstrates exactly the type of thinking these firms hire for.
Research Idea 5: Statistical Arbitrage and Cointegration Stability
Study the stability of cointegration relationships in equity pairs over time. Test whether structural breaks in cointegration can be predicted using regime-switching models, macroeconomic indicators, or changes in sector correlations. Measure the economic significance of pairs trading strategies after accounting for transaction costs, borrow fees, and execution slippage across different market environments.
Pairs trading is well-studied, but the question of why and when cointegration breaks down remains open. A paper that rigorously addresses this with modern econometric methods is publishable and shows you can think critically about strategy robustness rather than just curve-fitting.
Research Idea 6: Alternative Data Signal Decay and Crowding
Take a well-known alternative data signal (sentiment, satellite imagery, web traffic) and study how its predictive power has changed over time as adoption has increased. Use event studies or portfolio sorts to measure alpha before and after the dataset became commercially available. Test whether signal decay correlates with the number of hedge fund subscribers or AUM tracking similar strategies.
This research addresses the fundamental question of whether alternative data edges persist. A paper documenting signal half-lives across multiple datasets with proper controls is valuable to both academia and industry. Submit to the Journal of Financial Data Science or present at QuantMinds.
Research Idea 7: Reinforcement Learning for Optimal Execution
Apply reinforcement learning to the optimal execution problem and compare against classical Almgren-Chriss solutions. Design a realistic market simulator with temporary and permanent price impact, stochastic volatility, and variable liquidity. Rigorously benchmark the RL agent against TWAP, VWAP, and IS benchmarks across market regimes. Focus on whether RL genuinely outperforms after accounting for overfitting to the simulator.
RL in finance is overhyped but the execution problem is one area where it has legitimate potential. A careful paper that honestly evaluates when RL helps and when it does not is more valuable than another claim of superhuman trading performance. Target venues like the ICAIF conference or Machine Learning in Finance workshops.
Research Idea 8: Cross-Asset Execution Quality and Venue Analysis
Study execution quality across different venues, asset classes, or time periods using publicly available trade reporting data (TRACE for bonds, TAQ for equities, crypto exchange APIs). Measure effective spreads, price improvement, and information leakage. Test whether venue fragmentation helps or hurts execution quality for different order sizes.
Execution quality research has direct regulatory and commercial relevance. A well-executed empirical study is publishable in the Journal of Trading or Financial Analysts Journal and demonstrates the exact analytical skills execution desks need.
How to Maximise the Signal
The goal is independent validation. A project on GitHub is a weak signal. A paper accepted at a peer-reviewed venue is a strong one. To move up that spectrum:
- Write up your work as a proper research paper with abstract, methodology, results, and limitations
- Submit to relevant journals or conferences (Journal of Financial Data Science, ICAIF, QuantMinds, SSRN as a preprint)
- If publication is not feasible, present at university seminars or post to SSRN with a faculty co-author
- Push reproducible code to GitHub alongside the paper
- Be ready to defend every methodological choice under questioning
The single biggest mistake candidates make with projects is overstating their importance relative to fundamentals. Nail your GPA, pursue competitive math or programming achievements, secure internships, and publish research if you can. Projects are a supplement, not a substitute. If you do build them, be ready to explain every decision and line of code — interviewers will probe to see if the work is genuinely yours.
Which Research Direction to Pick
Match projects to the firms you target. For prop trading firms and market makers (Jane Street, Optiver, IMC, SIG, Jump), the options market maker and LOB microstructure projects hit hardest. For quant hedge funds (Two Sigma, DE Shaw, AQR, Point72 Cubist), the factor model, backtester, and alternative data projects land better. For HFT and execution roles (HRT, Tower, Virtu), LOB reconstruction and execution algorithm work carries most weight.
Pick two or three projects across a range, execute them carefully, and write them up honestly. Browse our quant job listings to see which firms are hiring and target your project work accordingly. Then practice with our interview question bank to be ready when the recruiter emails land.