Show Day
Personal project · in progressA fan-to-fan web app for concert day: resell spare tickets at face value or less, swap seats, find concert buddies and trade merch. Works in English, Traditional Chinese and Korean.
Concerts, crypto, everyday payments: when something matters to me, I turn it into a tool. I'm a FinTech graduate from Hong Kong, now in London, moving from payments operations into data analytics.
Question: does weighting tweets by engagement (how widely each tweet was shared) change the daily Bitcoin sentiment signal, or is a simple average enough?
Pearson correlation between the engagement-weighted and the simple (unweighted) daily sentiment score, n = 245 days, p < 0.001. They move together, but far from perfectly: weighting changes the signal.
Average weighted vs simple sentiment. Widely shared tweets were a little more positive than the average tweet.
Days with more tweets had slightly lower weighted sentiment. Busy days leaned more negative.
of days had positive weighted sentiment. Outliers (more than 2 standard deviations from the mean) ranged from −0.18 on 2019-04-14 to 0.75 on 2022-07-27.


Cleaned and deduplicated the tweets, scored them with VADER, built engagement-weighted daily scores, then checked the results again with a SQL re-analysis. News headlines were scored separately with FinBERT.
The 756,401 analysed tweets cover 245 days, mostly April to December 2019 plus a short window in mid-2022, so the two periods are not directly comparable. Sentiment here describes the conversation, not price moves. Daily values are not fully independent, so the p-value is a rough guide.
I've seen payments and crypto from the inside. At RD Technologies and KPay I worked on customer onboarding, KYC, AML checks and transaction monitoring, so I know how messy real financial data can be.
Outside work I build things: a sentiment pipeline for Bitcoin and Ethereum, a live crypto market data pipeline, a tool that compares what UK payment options really cost, and Show Day, a fan-to-fan ticket board for concert day.
Next, I want a data analyst role in FinTech, payments or crypto, turning messy financial data into clear answers a team can act on.
A fan-to-fan web app for concert day: resell spare tickets at face value or less, swap seats, find concert buddies and trade merch. Works in English, Traditional Chinese and Korean.
A sentiment pipeline for Bitcoin and Ethereum using VADER for tweets and FinBERT for news headlines, with weighted daily scores, portfolio weights and a Streamlit dashboard.
Live WebSocket feeds from Binance, Coinbase and Kraken saved into SQLite, with reconnect handling and cross-venue price spreads in basis points.
Compares UK payment options (debit, credit card, 0% card, BNPL and retailer finance) by expected cost, including slip risk, Section 75 and chargeback protection.
A decentralised ticketing app on Ethereum where each ticket is a non-transferable NFT tied to the buyer's wallet, to prevent fraud.

Tested an AI chatbot for HK MTR and found a gap in its restricted-words list. My testing feedback went into AI training.





I'm looking for entry-level data analyst roles in FinTech, payments and crypto in London. Email me at tamwinki0411@gmail.com or message me on LinkedIn.