Robert E. Hatem Email & Phone Number

Boston, MA, US

Boston, MA, US

* Research investment ideas using quantitative factors based on behavioral and risk-based rationale.* Build computation pipelines in R using in-house pipeline package and distributed computing.* Translated production codebase to Spark.

London, GB

• Downloading 8K reports, cleaning text, and representing as document-term matrices.
• Using topic modelling to compute a distribution over topics for each document (tf-idf, bag-of-words), perform dimensionality reduction using principal components analysis and latent direchlet allocation, then predict.

Boston, MA, US

• On a team with $35B assets under management, I improved one existing alpha factor by adding new data, 2B rows, from a data vendor. My research raised risk-adjusted returns enough that the team will implement the change.
• Clustered several dozen alpha factors using K-means, PAM, and DBSCAN algorithms using Python, R, and SQL. This aimed to fit models within clusters separately to achieve higher correlations, instead of all together.
• Researched new NLP sentence and document embeddings (doc2vec, SentEval) for a new alpha factor.

Wood Dale, Illinois, US

• As a one-person team, used random forests in Python to classify words from insurance cards as “member ID” or not in Python and Scikit-Learn. Used to automatically pre-populate forms after a patient scans their.
• The proof-of-concept classifier achieved a precision of 0.93 and a recall of 0.94.
• Parsed XML files to extract text from 20 insurance cards, engineered nine features based on simple text analysis (e.g., the fraction of characters in a string that is letters), and then binary classified strings as.

Chicago, IL, US

• Collected online web traffic data from 20 companies with seven monthly features collected over one and a half years using Python (Pandas, Scikit-Learn) to build a trading algorithm.
• Predicted next-month returns using random forests. Led a team of five students.

Madrid, Madrid, ES

• As a quantitative consultant, I wrote eight fifty-page reports subject to FR11-7 guidance on machine learning models used in consumer banking. The reports identified models risks; which regimes the model would stop.
• The eight models covered retail banking, marketing, and credit cards: household wealth prediction (linear regression), prediction of ATM cash drawdown by month (linear regression), prediction of credit card defaults.

Salt Lake City, UT, US

I taught two discussion sections of 30 students each for an introductory physics class. Classes met twice a week for 50 minutes and for 2 hours of office hours.

Liverpool Street, England, GB

Modeled daily dispatch of 536 power plants over 1,200 day historical period by looking at daily bids, consisting of 30 variables. Improved accuracy of estimating daily start cost by 40% - the company implemented the algorithm. Used fundamental data to predict electricity price structure to help asset managers maximize performance.

Chestnut Hill, MA, US

Collaborated with head coach to handle logistics so athletes can focus on training. 20 days of racing, about 30 days of training, 30+ hours per week commitment in-season. Organized $5000+ worth of training equipment for travel to races.

Chestnut Hill, MA, US

Simulated shining light on 5 micron long waveguides using simulation software. Work included in published paper. Worked under guidance of a professor and with a couple other students to create a better microscope.

Clemson, SC, US

Modeled fluid flow, derived equations and predicted in MATLAB, then compared to data. Recorded 10,000 data points from a video clip of a drinking moth to graph fluid flow versus time. Ruled out an elliptical shape of the proboscis, suggested more drinking mechanisms.

Described safe use of rental boat use to customers; pumped gas for recreational boats on Lake Winnipesaukee; provided cordial, prompt care to boats through vacuuming, wiping and waxing.

Master'S Degree, Financial Mathematics

Bachelor’S Degree, Physics

Marine Biology