HKPUG #101 - Behind the Python Call

Date: 19 September 2026 (Sat)
Time: 15:00–18:00 HKT
Coordinator: Alex Au

Please bring along your laptop and charger to code along! No GPU or powerful CPU is required.

Please fill in the form for admission: Google Form

Topics

Full-text Search Inside Your Python Process (Matthew, 15 minutes)

Do we need a vector database—or even a search server—for every search problem?

Matthew will compare several Python-accessible full-text-search choices before concentrating on Whoosh3: an embedded search library that keeps the index and ranking logic inside a Python application.

• See why BM25/BM25F remains a useful search baseline
• Index and query a small corpus with Whoosh3
• Write custom ranking logic in ordinary Python
• Compare the trade-offs with SQLite FTS5, tantivy-py, PyLucene and OpenSearch
• See where a small, read-mostly Whoosh3 index can fit inside a Python Lambda—and where that pattern stops scaling

Inside tinygrad: GPU Programming from Python (Qazalin, Talk + Workshop)

What happens after Python hands a tensor workload to the machine? Qazalin will introduce GPU programming with tinygrad, then lead a beginner-friendly hands-on tutorial.

The required exercises will have a CPU-compatible path in an organizer-provided VS Code Web environment. Qazalin has confirmed that any ordinary CPU is sufficient. Basic Python is enough; no prior AI, machine-learning or GPU-programming experience is required.

Participants may optionally use a compatible local GPU if the final tested setup supports it, but the required workshop path does not depend on one.

Capacity
80 participants (provisional, pending venue confirmation)

Venue Info
City University of Hong Kong — exact room TBC.

Rundown

• 15:00–15:20 Registration and networking
• 15:20–15:25 Upcoming events and community announcements
• 15:25–15:30 Opik challenge winning-team award presentation
• 15:30–15:45 Whoosh3, BM25F and custom scoring — Matthew
• 15:45–16:15 Inside tinygrad: GPU programming from Python — Qazalin
• 16:15–16:25 Break and workshop setup
• 16:25–17:45 tinygrad hands-on workshop — (CPU or GPU)
• 17:45–18:00 Workshop debrief, Q&A and closing
• 18:00 onwards Networking outside the venue

Speakers Bio

Matthew
Matthew is a Python practitioner who previously worked at Hong Kong Economic Times. In this short session, he will focus on the practical engineering choices behind embedded full-text search and hackable Python ranking.
LinkedIn: Si Heng Matthew Yuen

Qazalin
Qazalin will introduce tinygrad and lead a beginner-friendly GPU-programming tutorial. A fuller speaker biography and project links will be added from the speaker's final copy.
LinkedIn: Qazal Samani

Audience Pre-requisite

• Beginner-level Python knowledge is sufficient
• No prior AI, machine-learning, GPU-programming or CUDA experience is required
• Bring a laptop, charger and a current web browser
• Any ordinary CPU is sufficient for the required workshop path
• A compatible local GPU is optional, not required

If you cannot bring a laptop, tell the organizer in your RSVP answer.

How to join?

Fill in the form for admission: Google Form

Click Attend on this page

Venue access instructions, the VS Code Web access details, workshop repository and pre-flight check will be shared with confirmed attendees before the event.
This is a free community meetup. Please RSVP only if you plan to attend.

Final workshop exercises will be confirmed before publication.