Work engagements


Credit Suisse // Advanced Execution Services (AES), High-Touch Sales Trading Projects include:

  • BC&E cost analysis for systematic traders (analyzed execution patterns of clients)
  • Made one-pagers for market microstructure analysis post-crash
  • Worked on cross-trade opportunity detection tool
  • Blocks sheet spreadsheet that summarizes all cross trades for most crossed stocks
  • Block pricing model based on modifications to the Algrem Chriss model
  • Execution optimization model for the program trading desk (how to execute different orders while maintaining natural hedge)
  • TCA analysis
  • cross-border trade settlement escalation pathway mapping

Société Générale Equity Derivatives Intern

  • Developed quant algorithms with IT, Trading, and Financial Engineering teams
  • Accomplishments include: building automated trading system using C#, building an autoML (automated machine learning) library, deriving novel correlation model from gravitational black hole theory, adopting Hidden Markov Models to detect hidden states, deriving algorithmic game theory from scratch and applying them to bid-ask level 2 financial data, modelling bid-ask statistical relation using gravitational force relations, conducting Monte Carlos for risk modelling of insurance data for Guaranteed Minimum Benefits Products (GMxB), developing generative adversarial networks to speed up simulation process of Monte Carlo, enhancing risk modelling for banks by building enhanced Neural Net – Monte Carlo (NN-MC) hybrid algorithms, building political arbitrage algorithms to test for market inefficiencies from market news release via Twitter using sentiment analysis, performing actuarial variable annuities modelling of insurance death data by creating own models for risk classification, building options pricing models with HSI warrant data
  • built quant models using gravitational theories, Markov chains, algorithmic game theory, Monte Carlos, generative adversarial networks
  • worked on trading systems, auto ML library, options pricing, political arbitrage model, insurance policy & variable annuity models
  • worked on test driven development, building GUI functions, topological data analysis

Machine Learning Research Assistant The Hong Kong University of Science and Technology Sept 2019 - Feb/Aug 2020 | Worked on grammar error correction system with novel deep learning architecture in collaboration with Language Centre and Information Systems department. | Developed new adversarial attack algorithms against text classification models & corresponding unified defenses | Submission to ACL2020, ICML2020

Research Assistant, Department of Management HKUST Business School Feb 2018 - Aug 2018

  • Automated data collection processes for Management department; professor’s research project related to diversity on work platforms (gender, race, ethnicity, …)
  • Built a web-scraper, and built an AI solution to bypass CAPTCHA security mechanisms
  • Managed room of 11 computing resources simutaneously
  • Analyzed data, plotted and studied social network interaction charts and relations within employees and employers

Venture Capital intern The Aria Group Jan 2018 - Feb 2018 • Proved profitability of SwapIt Ltd, an Aria Group portfolio company that was underperforming, by applying data science fundamentals and performing rigorous data analysis • Analysis and research helped founders raise 172,000 HK$ investment, allowing the company to operate for at least 2 more months

Intra-day Trader 8Vesting Nov 2017 - Feb 2018 • Used volume, bands, moving averages to day-trade night-shift on US Equities • Average stats: 84 trades; US$ 62 P&L; 566,775 Volume (per night) Learning to Trade.jfif

Lead Technology Advisor Coxana Innovation | INSPIRE Education/Studio May - July 2019 Developed swarming & image detection software for Tello EDU drones; built a curriculum for high school students to learn drone programming

Data Scientist Surepromise June-July 2017

  • Determined the search algorithm of eBay and Amazon thanks to Big Data analytics and machine learning algorithms (search engine optimization & ranking algorithm cracking)
  • Conducted Business Process Reengineering to optimize development of their e-commerce platform
  • Crunched data to suggest actionable decisions regarding items listings design, advertising strategy, search engine optimization in European markets
  • Built an e-commerce platform along with various analysis and aggregation algorithms on backend
  • Requirements Reprioritization Requirements Reprioritization.jfif Part of the process done for BPR (Business Process Reengineering) - basically I analyzed each feature required for their platform and determined its strategic value by considering various factors then presenting the results as this graph
  • Web-scraping Web-scraping.jfif One of the web-crawlers designed to scrape through Amazon for comprehensive item information.

Developer & Founder BreakupBot Nov 2016 - May 2017

  • Built an AI therapeutic chatbot that empathetically converges from psychological counseling to philosophical counseling
  • Conducted analysis on human psychology and personality types before training the bot to respond appropriately to specific personality traits
  • Hypothesized, built and tested Heartrate detection via Camera-LED to deduce stress level with the end-result of mimicking empathy
  • Possesses an active user base currently The bot uses your heartrate (detected by Camera LED) to alter the conversation tone immediately.

HKU Faculty of Law Unison

Sales And Marketing Intern New World Development Company Limited Central, Hong Kong

  • Developed algorithms to predict time series housing property prices (artificial neural networks, genetic algorithms, probabilistic modeling, scenario analysis, etc)
  • Designed a prediction tool to foretell the outcomes and consequences of any decision
  • Applied sentiment analysis to obtain marketing insight for sales (news source bias, market sentiment over time, trends, etc)
  • Built a chatbot for online customer registration & an internal departmental chatbot to extract sales data from databases, valued at HKD 212,000 Pitching ideas to Leaders Sidd’s Spirit - An Artisan’s Chatbot A chatbot I made to detail the work I did at NWD, and to prove the notion: a chatbot is a part of you. After all, when someone messages this chatbot at 3am at night, they will be talking to a part of me and my sentences. LSTM model testing ANN Pitch The company had been used to using Excel modeling and qualitative analysis until I not only predicted property prices with over 90% accuracy, but also pitched ANN’s value to Heads of the company.

Equity Research Analyst Invest Bots Limited Central, Hong Kong Dec 2016 - Feb 2018 • Performed extensive financial analysis including DCF, financials, fundamentals, sensitivity analysis, scenario analysis, algorithmic stock price projections, and compiled all the results into spreadsheet models for fund managers investing in the airline, shipping, and electronics sectors Late-night Financial Modelling Being new to the team means putting in more effort to climb the ropes faster - that’s what it means to be an analyst. Tips from the Head of Investor Relations of Tencent

  • Head of Investor Relations & Global Communications (Catherine Chan) at Tencent led the discussion on being a better analyst.
  • She shared tips, ranging from analyzing publicly-available information efficiently, to understanding numbers and the story behind, and what-not.

Research Assistant, Language and Cognitive Development Lab University of California, Berkeley Sep 2018 - Feb 2019 Berkeley, California

  • Building annotation tool to label content words in dialogues of children to study their language use

Engineer, CalUnmanned University of California, Berkeley Nov 2018 - Feb 2019 Working on autonomous drones, developing flight paths using reinforcement learning


  • HKUST Business School 2016-2020 Traders@UST | Publishing work at | Presented a paper on a modified therapy to combat HIV drug resistance by predicting mutational pathways of a given strain of HIV in a patient | Developed a complete trading algorithm under 9 hours, for a 10-year long-term investing strategy of investing in government treasury bonds and generating annualized returns of ~10.4% | Built a team cohesiveness algorithm to estimate team success/performance based on personality analysis | Used sentiment analysis algorithm to quantify market share of competing products and determine product market sentiment

Fixed Income Return Maximization.jfif Built the entire implementation under 9-hours, including data cleaning, normalization, converting to stationary, performing additive regression, bond regression, and portfolio weighting optimization. Data Mining with Python.jfif Through the use of Python, one can apply statistical algorithms including regressions and what-not to automate the computation of a vast amount of data. In this particular snapshot, the code details how one would divide the original dataset into training and testing data, which would be used to compare whether the model built is consistent or not.

UC Berkeley 2018-2019

Stanford 2017-2017

La Salle College


Hong Kong Unison Sep 2016 - Present

  • Presently one of 20 students sponsored for academic study and advocacy work
  • Advocated for equal opportunities for ethnic minorities in education
  • Spoke at the Legislative Council; attended meetings with Chief Executive candidates; received training in advocacy, policy address drafting, political ideologies
  • Delivered talks at secondary schools on interview skills and promoting university education; conducted surveys with kindergarten parents
  • Prior projects included: starting a mentorship programme for ethnic minority F.6 students; building a chatbot to ask questions in Math & Chemistry, and autonomously grade student’s responses

Time Auction Sep 2017 - Dec 2017 Provided recommendations for future growth as part of the GBUS2020 course

Undergraduate Hall VIII & IX Jan 2017 - Jan 2018

Test scores

DELF The DELF and DALF are diplomas awarded by the French Ministry of Education to prove the French-language skills of non-French candidates.

HKDSE 32/35 • Top 1% of candidates in Hong Kong • Highest grade (5*) in English Language, French Language, and Chinese Language for Non-Chinese Speaking students • Second-highest grade (5) in Mathematics, Mathematics Extended Module 1 (Calculus and Statistics)


  • A test of verbal reasoning skills
  • A test of my ability to understand and interpret information
  • A test of inductive and deductive reasoning abilities
  • A test of my ability to analyse information and draw conclusions

The attained score is higher than the average of 23/42*; the exam indicates the presence of a legal mindset.

*There is no pass mark for the exam


DeepObfusCode: Source Code Obfuscation Through Sequence-to-Sequence Networks arXiv The paper explores a novel methodology in source code obfuscation through the application of text-based recurrent neural network (RNN) encoder-decoder models in ciphertext generation and key generation. Sequence-to-sequence models are incorporated into the model architecture to generate obfuscated code, generate the deobfuscation key, and live execution. Quantitative benchmark comparison to existing obfuscation methods indicate significant improvement in stealth and execution cost for the proposed solution, and experiments regarding the model’s properties yield positive results regarding its character variation, dissimilarity to the original codebase, and consistent length of obfuscated code.

HAIV: Computational Precision Medicine against HIV Drug Resistance Feb 23 2018 Human Immunodeficiency Virus (HIV) is currently a non-curable yet treatable virus. The virus replicates by encoding its own DNA into white blood cells, and the virus prevents white blood cells from protecting the body from other diseases, thus causing immunological failure. The most effective treatment currently in use is antiretroviral therapy (ART) (commonly known as “cocktail therapy”), which involves prescribing to the patient a combination of drugs that have a high likelihood of inhibiting HIV replication. In brief, the solution is to forecast the development of HIV from strain to strain, by predicting the mutations it undergoes, which will help us apply the right drug at the right time. In more detail, we have an algorithmic methodology to predict behavior of HIV in the human body with respect to current, historical, and future conditions of the virus, and with respect to the human body conditions and medication applied over time. The second stage is to build the machine that automatically detects HIV strain types, and applies the medication. The longer term goal from this solution is to lay a foundation in precision medicine, in particular for anti-drug-resistant viral medication treatment. To summarize, this paper aims to put the AI in HIV.

Mental Approximation Processing Micorsoft ImagineHack 2018 Jan 20 2018

  • Developed the backbone of an auditory Google Glass to provide mental, emotional, performance, and ethical advice on the fly.
  • Built ethical and relational AI analysis from scratch for AI (Ethics Engine, and Relational Analysis)

BreakupBot: An Artificial Intelligence Therapeutic Chatbot Conducting Empathetic Counselling via Heartrate Sentimental Analysis 10 May 2017 A research paper describing the market need for a novel therapy method, the design and methodology behind BreakupBot, and the results following the release of the chatbot.


ETHGlobal: HackMoney - Winner Winner in pToken category



Marshall International Case Competition (Top 8) Top 8 out of 30 teams; proceeded to Challenger round

HKUST Bizkathon - 1st runner-up Built privacy-preserving system based on homomorphic encryption for virtual banks to obtain data from customers and perform machine learning or analytics on the data while still maintaining privacy of customers.

source code:

CalHacks 5.0 - Winner (Epson Sponsor Award) Built AR goggles that helps Alzheimer’s patients find lost objects using object detection.

SacHacks 2018 - Winner (Smartcar 1st Runnerup, Best First-Time Hack, Best Domain Name, 3rd Overall) We built a centralized system for cars to use cameras to detect crime and notify authorities. It uses object detection to detect weapons, takes snapshots, sends messages to 911 via Twilio, and updates a dashboard of crime using the cars’ built-in geolocation tagging mechanisms.

Video link:

Slide deck link:

AngelHack 2018 - Blockchain Category - Champion Built complete dAPP on WAVE platform with data transactions and smart contracts for placing sports bets with escrow to manage counterparty risk of betters and bookkeepers.

White paper: Repo:

Access to Justice InnoTech Law Hackathon 2018 - Honorable Mentions Pitched and justified the development of a dynamic UX integrated application platform for government legal aid

Classified Post Hackathon 2018 - Champion Hard-sold and pitched a headline recommendation engine for editors of South China Morning Post

CityU Hackathon 2018 - Most Innovative Idea & 1st Runnerup Built an automatic lecture transcription tool that also summarizes and quizzes students based on what was taught.

HKUST-Radica Big Datathon - 2nd

  • Built a Youtube-scraper that analyzes people and objects in video footage to extrapolate fashion trends

Cyberport Creative Micro Fund Won HKD 100,000 in under 5 minutes by pitching a startup to a panel of investors at Cyberport

EasyShip 24 Hour E-Commerce Hackathon 2017 - 1st Runner up Built a revolutionary UX-centric platform to tailor easyship’s UI to the needs of crowdfundees through prediction analytics

Hack the Runway 2017 - CHAMPION Inspiree - we built a platform that autotags fashion images uploaded by users with image classification and NLP, with a dashboard to present interpretable insight

HardUST 2017 - 1st Runner-up Worked in a team of four to develop a novel system that caters to autonomous vehicles by standardizing a unique design among all vehicles to permit free-flow throughout the system with absolutely no delay. Our team built prototypes to demonstrate how the custom vehicles integrate and flow through the system.

Particular contributions include:

  • Giving the winning pitch to rope in investors/judges
  • Designed and created all the animations for demonstration purposes using Adobe Flash
  • Brainstormed and scrutinized the to-be-built system and mechanisms
  • Sustained the team by bringing hardware materials and food regularly
  • Modelled the two prototypes that we built using 3D modelling software, including Google Sketchup and Blender
  • Specialized in designing and printing the parts needed using 3D printer

Joseph Lau Luen Hung Charitable Trust Scholarship Hong Kong Award An HKD80,000 p.a. renewable scholarship for students who care for people, so much so that awardees are typically multilingual and are dedicated to community service.

Br Bernard Guellec Memorial Prize for French An annual award for the top performing students in French in the Hong Kong Diploma of Secondary Education Examination (HKDSE).

University Entrance Scholarship In recognition of outstanding performance in the Hong Kong Diploma of Secondary Education Examination (HKDSE)

Harmony Scholarship (2016) Awarded for demonstrating efforts to progress racial harmony, notably for participation in ethnic-minority-related activities, through aiding foreign festivities (Durga Puja for Hindus on behalf of HKBA, Bangladesh Independence Day for Bengalis on behalf of HKBCC).

Fortune Pharmacal Scholarship The scholarship was given to me as the awarding orgainzation believed I showed passion and interest in the betterment of lives of those in need. Not only does the scholarship provide me short-term financial support, but it also acted as a platform for me to make a greater impact. For example, after becoming a scholar of Hong Kong Unison, this NGO allowed me to start my own projects with them, and thus giving me more freedom in how I wish to change lives, instead of simply following their preset activities. Thanks to the award, I can be the frontier in community service innovation.

Good Lasallian Award Scholarship For the Good Lasallian Award, it is granted by the Principal to those who demonstrate the values of a Lasallian in their lives. The values that my school believed I encompass include compassion, prudency, perseverance, humility, among others.

Br Felix Sheehan Memorial Prize for English An annual award for the top performing students in English in the Hong Kong Diploma of Secondary Education Examination (HKDSE).

French Subject Prize (2011-2015) To receive the subject scholarships, it is necessary to show outstanding performance in that subject. Coming First in Form in the respective subject is rewarded with the prize, as it probably shows commitment and dedication to perfecting one’s ability in the subject.