Hi, my name is

Zhouhan (Han)

I read Mathematics & Economics at LSE (Graduating 2028). Most of the time, I'm building cool things and working with data.

Download CV

About me

Most of what I build starts with something I need. My travel tech startup, Atlas, because I wanted to keep track and share my travels digitally, Accenture’s test automation software because my team needed to speed up testing. I enjoy using data to make what I build more effective, working with travel patterns, testing frequency, housing prices, etc.

I also make videos for LSE’s Instagram and YouTube channel (50k+ views), run marathons, and climb rocks.

Selected work

Atlas (Travel Tech Startup) | 100+ Users in Beta

Co-Founder · 2025-2026

Built Strava for travel, where users plan, track, and share trips on a 3D globe. Ran a closed beta amongst 100+ users in university hall, pitched at LBS startup event.

How it works

  • Travel Blend (like Spotify Blend, but for travel). Built on Jaccard and cosine similarity on a 10-dimensional travel vector
  • Travel Planning with real time travel durations and directions through Mapbox and Google Maps API
  • PostgreSQL/Supabase backend
  • PostHog engagement and behaviour pipelines
  • TypeScript
  • React
  • Next.js
  • PostgreSQL
  • PostHog
  • Figma
Live

London Property Price Model | Best Project Award

ML Engineer @LSE Data Science and Quant Society · 2025-2026

Built a London property price prediction model in a team of 4. Owned the model architecture and feature engineering, and stepped in as Project Manager when required. Achieved Best Project Award in the 2026 LSE Data Science and Quant Society Project Showcase.

How it works

  • LightGBM trained on 500k+ rows, 11% MAE, 91% R² on unseen data
  • Optuna hyperparameter tuning
  • Huber loss for robustness to outliers
  • Model segmentation from error analysis for 3 price segments
Interactive dashboard mapping predicted London property prices by postcode area, with filters and per-area market data Model-selection reasoning — a continuous price target with non-linear spatial interactions and heteroskedasticity leading to gradient-boosted LightGBM

Finance 2.0 + AI Chatbot | 20+ Hours Saved/Month

Resource & Budget Planning @Singapore Police Force · 2023–2025

My day-to-day at the Singapore Police Force involved managing our operational budget. I went beyond my job scope to build an automated budget management system using VBA and PowerBI, saving 20+ hours per month. Observing that my colleagues were answering repetitive queries, I integrated an AI Chatbot into our intranet, cutting query workload by >38%.

How it works

  • PowerBI + VBA automation replacing manual monthly reporting
  • GovTech VICA chatbot rollout on the intranet, customised to our values, tone, and knowledge
  • Random Forest classifier informing allocation of S$10M+ funds
  • PowerBI
  • VBA
  • Python
  • Random Forest

Test Automation Software | >80% Test Duration Saved @Accenture

Automation Engineer @Accenture · 2025

Observing that too much testing was done manually and repetitively, I wrote a Playwright program to automate UI interactions to speed up testing. Eventually expanded to a full-stack web app which allows non-technical team members to conduct tests efficiently.

How it works

  • Playwright for UI interactions in browser
  • Streamlit for frontend
  • Cross-team coordination in Azure DevOps and ServiceNow
  • Python
  • Playwright
  • Streamlit
  • pandas

Diplomacy & Financial Flow Model | Spotlight Scholar Award

Network Scientist @Cambridge Centre for International Research · 2024

Network science research on how countries’ diplomatic ties relate to financial market inflows. Nominated as CCIR Spotlight Scholar by mentor Dr Daniele Cassese of Cambridge University for outstanding quality of research.

How it works

  • Graph-theoretic analysis of the diplomatic network
  • Panel regressions on IMF capital-flow datasets
  • Network metrics computed and visualised in Gephi
Diplomatic representation network of 200+ countries, force-directed and coloured into four modularity classes by their diplomatic ties

Other builds

  • London Pub Insurance Platform Parametric insurance payouts on Flare, built at ETH Oxford Hackathon.

    JavaScript · Solidity GitHub

  • Paodekuai Tracker Score tracker for 跑得快 (paodekuai) card games my family plays every Chinese New Year.

    Python · Streamlit GitHub

  • LSE Math Problems MA103 abstract mathematics exercises solved alternatively using code.

    Python GitHub

03. What's next?

Chat with me

Looking for opportunities to apply data science to solve real world problems. Always up for a chat about ML, startups, or whatever you're building.

Say hello