TAHSIN JAWWAD · PORTFOLIO
TJTAHSIN JAWWADMQF · WATERLOOOPEN TO QUANT & SWE ROLES
MASTER OF QUANTITATIVE FINANCE · UNIVERSITY OF WATERLOO

TAHSIN
JAWWAD

Quant researcher & software engineer

Quantitative finance at Waterloo, computer science from UBC. I do the research, and I build the software it runs on.

OPEN TO
PythonC++SQLpandasstatsmodelsDocker
VIEW MY WORKRÉSUMÉTRANSCRIPT GITHUB LINKEDIN
4.32

GPA out of 4.33, BSc Computer Science at UBC

96.7%

Cumulative average, with 100% in time series, regression and machine learning

$110K

In scholarships, including the $80K International Major Entrance award

PROFILEWATERLOO, ON
Portrait of Tahsin Jawwad
TAHSIN JAWWAD
QUANT · SOFTWARE
NOW
MQF, University of Waterloo · Dec 2027
BEFORE
BSc Honours CS, UBC · GPA 4.32/4.33
STACK
Python · C++ · SQL · R · Java
UBC COURSEWORK · FINAL %
Time Series & Forecasting100
Applied Regression100
Machine Learning100
Stochastic Modelling & Simulation99
Matrix Algebra99
Sampling & Design99
Differential & Integral Calculus100
Analysis of Algorithms98
FULL TRANSCRIPT
HIGHLIGHTS
  • NOWMaster of Quantitative Finance at the University of Waterloo
  • GRADUATEDBSc Computer Science, UBC: 96.7% average, Dean's Scholar
  • THESISFirst-author manuscript on turning plain-English questions into MongoDB queries
  • RESEARCHCrypto statistical arbitrage, backtested on survivorship-free data
  • C++NASDAQ order book feed handler: 14M messages a second, 121 ns median
  • SCHOLARSHIPS$110K awarded at UBC
  • OLYMPIADGold medal, ranked 1st in the UAE
QUANT, RESEARCH & ENGINEERING

PROJECTS

QUANT RESEARCHPERSONAL PROJECT · 2026 · LIVE FORWARD TEST RUNNING

Statistical Arbitrage in Cryptocurrencies

Momentum, order-flow and funding-rate strategies on six years of Binance data, combined into one walk-forward portfolio. On a year of data it never saw during research, it returned a 1.46 Sharpe ratio with almost no exposure to Bitcoin. Then I tried to break it, and a revised version is now running a live forward test.

1.46

Sharpe ratio on the held-out year

−0.02

Beta to Bitcoin: returns that don't just ride the market

1.68 / 2.42

Funding-rate strategy Sharpe, development / validation

47

Strategy variations tested, all logged to correct for overfitting

PythonpandasNumPySciPystatsmodelsDeflated Sharpe
CODE REPORT
STRESS TESTHELD-OUT YEAR, JUL 2025 – JUL 2026

Sharpe ratio after each realism check

Original backtest+1.46
Rebuilt from exchange archive+1.45
Stablecoins & tokenized stocks removed+1.26
Delisted coins added back+0.35
Funding trade priced on futures−0.01
Limit orders must actually fill+0.12
−0.91

Survivorship bias. My original coin list only had coins that still exist today. Adding back the ones that died erased most of the returns, which is exactly what the test was for.

QUANT DEVPERSONAL PROJECT · C++20 · 2026

NASDAQ Feed Handler

Reads a full day of NASDAQ TotalView-ITCH market data and keeps a live limit order book for every stock. A simple reference book checks the fast one after every order event, and a second thread computes order flow imbalance from the book over a lock-free queue.

423M

Messages in the replayed trading day, across 8,915 symbols

121 ns

Median time to decode a message and update the book

0

Mismatches against the reference book, checked after every event

0

Heap allocations in the fast book; the reference made 296 million

  • Flat price arrays with a bitmap to find the next best price, a preallocated order pool and an open-addressing hash map, all set up front on huge pages.
  • Tested on Linux and Windows CI, including address, undefined-behaviour and thread sanitizer builds.
C++20CMakeLock-free queueSanitizersLinux & Windows
CODE
BENCHMARKNASDAQ SAMPLE DAY · INTEL i5-12400

Messages per second

Fast book14M
Reference (std::map)3.6M
About 4× the standard-library version

Time per message

Median (p50)121 ns
p90282 ns
p99552 ns
p99.9874 ns
Every one of the 423M messages timed on its own
HONOURS THESISFIRST-AUTHOR MANUSCRIPTUBC · 2026

Text-to-NoSQL

Asking a MongoDB database questions in plain English. A language model turns the question into SQL, and my translator turns that SQL into MongoDB aggregation pipelines that run inside the database, instead of pulling every record out and filtering in Java.

  1. Questionplain English
  2. LLMDAIL-SQL writes SQL
  3. My translatorSQL → MongoDB
  4. MongoDBruns natively
  • Added native support for joins, GROUP BY / HAVING, UNION / INTERSECT / EXCEPT, correlated and scalar subqueries and computed columns, plus clean-up passes for SQL quirks that LLMs produce.
  • End to end it answers 88.4% of questions correctly, ahead of prompting Gemini directly (86.8%) and of GPT-5.6 given the same SQL (81.8%), with no extra LLM call once the SQL exists.
JavaPythonMongoDBLLMsDocker
CODE THESIS (PDF)
RESULTSTEND-SPIDER · 2,775 QUESTIONS

Questions answered correctly

My translator, correct SQL96.0%
My full pipeline88.4%
Gemini, direct86.8%
GPT-5.6, same SQL81.8%
Results checked by running each query on MongoDB

Questions translated successfully

Translator I started from63.2%
After my changes98.9%
1,021 failures down to 32, zero regressions
SIDE PROJECT2026

STAB Lab

Desktop team and battle lab for competitive Pokémon Showdown: team builder, damage calculator, replay review and practice games against three levels of bot, plus a matchup model trained in Python and run in the app with ONNX.

ElectronReactTypeScriptPythonONNX
CODE ENGINEERING
WHERE I'VE STUDIED AND WORKED

BACKGROUND

EDUCATION

  1. SEP 2026 — DEC 2027IN PROGRESS

    Master of Quantitative Finance

    University of Waterloo
    • Stochastic Calculus for Quantitative Finance, Estimation & Hypothesis Testing.
  2. SEP 2022 — APR 2026

    BSc Honours, Computer Science

    University of British Columbia · Minor in Data Science
    • 96.7% average, GPA 4.32/4.33.
    • Dean's List 2022–24, Dean's Scholar 2024–25, Faculty of Science International Student Award.
    • $110K in scholarships, including the $80K International Major Entrance Scholarship.
    • View transcript
  3. 2022

    A Levels

    The Winchester School, Jebel Ali · Dubai
    • 4 A*, Best Across Three in the UAE.

EXPERIENCE

  1. MAY — DEC 2025CO-OP

    Junior Developer

    Aether Automation Inc. · Vancouver
    • Automated Zoho workflows in Deluge, Python and JavaScript that sync data across teams, cutting manual work by 70%.
    • Built an OpenAI-based classifier that sorts customer emails and drafts replies, cutting response time by 60%.
  2. 2025 — 2026RESEARCH

    Honours Thesis Researcher

    University of British Columbia
    • Text-to-NoSQL via SQL, written up as a first-author manuscript. Details in Projects above.
LANGUAGES, METHODS, TOOLS

TOOLKIT

WHAT I WRITE

Languages

  • Python
  • C++
  • SQL
  • R
  • Java
  • TypeScript
HOW I TEST IDEAS

Quant methods

  • Backtesting
  • Time series & regression
  • Monte Carlo simulation
  • Trading-cost modelling
  • Overfitting checks
  • Stochastic calculus

Stochastic calculus is part of my MQF coursework this term.

DATA & ML

Libraries

  • pandas
  • NumPy
  • SciPy
  • statsmodels
  • scikit-learn
  • PyTorch
SHIPPING IT

Tools

  • Git
  • GitHub Actions
  • Docker
  • Jupyter
  • MongoDB
  • MySQL
AWARDS & COMPETITIONS

HONOURS

Gold #1 UAE

Mathematics Olympiad

International Olympiad Foundation · UAE Rank 1 in 2019, Rank 3 in 2020

3 years

Dean's List & Dean's Scholar

University of British Columbia · Dean's List 2022–24, Dean's Scholar 2024–25

$110K

Scholarships

University of British Columbia · including the $80K International Major Entrance Scholarship

2019

UAE Team, 16th Hanoi Open Mathematics Competition

Hanoi, Vietnam

2022

Best Across Three in the UAE

Cambridge International A Levels · 4 A*

05 · CONTACT

LET'S TALK

Hiring for quant research, trading or software engineering? I'd like to hear about it.

TAHSIN.JAWWAD23@GMAIL.COMLINKEDINGITHUB