tech-econ

The largest curated library for tech economists

Packages, datasets, papers, and learning resources

Coverage

527

Packages

443

Datasets

1377

Papers

518

Resources

Curated from 500+ sources across causal inference, experimentation, pricing, ML, and econometrics

Hybrid Search Architecture

Lexical

MiniSearch (BM25)
client-side, ~10ms

Semantic

GTE-Small ONNX
384-dim, quantized

↓

Reciprocal Rank Fusion

score(d) = Σ 1/(k + rank_i(d)), k=60

↓

Re-ranking

score × (1 + 0.4 × model_score)

Latency <50ms total (client-side)

Learning-to-Rank Model

Target Variable

y = 5·clicks + 1·dwell + 2·scroll + 0.5·impressions

↓

Categorical

18 features
section, type, level...

Semantic

384-dim embeddings
GTE-Small

↓

LightGBM Tweedie

100 trees · max_depth=5 · p=1.5

↓

Warm

Observed engagement

Cold

Predicted × 0.3

Recommendation System

Item Embeddings

bge-large-en-v1.5 · 1024 dimensions

↓

Similarity

sim(a,b) = (a·b) / (‖a‖·‖b‖)

↓

Related Items

Top-5 neighbors
threshold > 0.3

For You

History + favorites
recency-weighted

Storage Client-side (localStorage)

System Architecture


Static Site:     Hugo → Cloudflare Pages (275+ PoPs)
Search Index:    MiniSearch (BM25) + ONNX Runtime
Embeddings:      bge-large-en-v1.5 (1024d) / GTE-Small (384d)
Ranking Model:   LightGBM Tweedie (offline trained)
Analytics:       Cloudflare Workers + D1 SQLite
Privacy:         No cookies, DoNotTrack respected

3130 items · 527 packages · 1377 papers · 443 datasets