Bet Guasch
03ProjectData Science / Music / Prototype01 — 08

Set
Builder

A decision-support tool
for DJ set planning.

Black and white turntable image for the Set Builder project

A data-driven tool that helps DJs design coherent sets around anchor tracks, energy progression, mood and musical compatibility.

View prototype →Read case study

01 / Input

Set length
Anchor tracks
Mood & energy arc
Preferences

You set the direction.

02 / Engine

Track similarity
BPM & key compatibility
Mood analysis
Transition scoring
Sequence optimisation

Data finds the possibilities.

03 / Output

Suggested tracklist
Energy curve
Transition insights
Alternative paths

You make it yours.

Build your set

◼ Romy — Strong×
◼ Fred again.. — Delilah×
◼ Jamie xx — Gosh×
Energy journey

Suggested 60-min set

Built around 3 anchor tracks · 12 tracks · 94% coherence

Illustrative
#TrackBPMKeyMoodTime
01Kiasmos — Looped1228AAtmospheric4:48
02Bicep — Glue1238AWarm5:12
03Romy — Strong1238AUplifting5:06
04HAAi — Satellite1249ADriving4:50
05Fred again.. — Delilah1249AEuphoric4:42
06Mall Grab — I’ve Always...1259AEnergetic4:58
07Overmono — Good Lies1269AHypnotic5:10
08Jamie xx — Gosh1249ATension5:02
09Chloé Caillet — Transient12510APeak4:36
10Barry Can’t Swim — Suns...1239AWarm5:08
11Elkka — Harm1228ADreamy4:54
12FKA twigs — Jealousy1208AEmotional5:02
59m 48sTotal time
123.8Avg BPM
8.7/10Avg transition score
9Peak at track
Prototype statusCurrent stage — exploratory prototype

Interface, track sequence and scores are currently illustrative. The next phase connects the prototype to public music datasets and a real scoring model.

Next → Data ingestion · Feature engineering · Transition model

Questions explored

  1. 01

    Can data help build smoother transitions between tracks?

  2. 02

    Can mood and energy be approximated using audio features and tags?

  3. 03

    Can a set be optimised around anchor tracks and duration constraints?

Data & methodology

Data sources

Million Song Dataset (subset)
Last.fm tags & similarity
Open audio feature datasets
Genre metadata

Methods

Feature engineering
Similarity scoring
Clustering & embeddings
Sequence optimisation

Tools

Python · Pandas · Scikit-learn
Jupyter Notebook
Data visualisation
Astro prototype

Key learnings

Data can suggest structure, reveal connections and make exploration faster. But a great set still comes from taste, intuition and context. This tool is a starting point for better decisions, not a replacement for the creative process.

“Algorithms can find the path.
You choose the journey.”
This project demonstrates

Data Science · Product Thinking · UX · Creative Problem Solving

← Back to projects