Research-grade AI pipelines for universities and businesses — patrickchu.net
Re-run the deposited data and scripts to inspect 15 reproduced comparisons and four transparently labelled close results.
Open demo →Re-run the deposited word norms and compare crowdsourced with AI-generated estimates, while marking analyses that require unavailable external data.
Open demo →Re-run the deposited acoustic data and compare the reported group-level patterns, while keeping the small sample and omitted post-hoc tests visible.
Open demo →Compare the archived experiments with the paper's core age-perception results, while showing the omitted sub-models and effect-size convention.
Open demo →Test three human-reviewed administrative workflows on fictional materials: email triage, reference-letter drafting, and meeting-note summarisation.
Open demo →Re-run two deposited experiments and inspect whether the reported statistical pattern survives modest resampling variation.
Open demo →A live pipeline checks the archived survey data against the paper—reproducing the available results while showing why the headline before/after analysis
Open demo →Compare nine deposited-data tests across four studies, including the substantially smaller Hong Kong effect and software-related numerical differences.
Open demo →Re-run the deposited pretest correlations and regressions for 35 children, while clearly separating statistical agreement from generalisability or
Open demo →Compare three AI-generated drafts with published Chinese items to inspect the workflow—not to validate a finished bilingual instrument.
Open demo →See how a short synthetic lesson moves from audio to an illustrative coded transcript.
Open demo →Explore how a real single-subject EEG recording is cleaned, averaged, and visualised—and where a tutorial workflow stops short of study-level evidence.
Open demo →Follow a transparent public-data EEG pipeline from raw signals to group statistics, with five effects clearing the pre-specified gate and two retained as
Open demo →Check five published eye-movement measures and two reading effects in the Provo Corpus complete-case subset.
Open demo →Trace a synthetic reaction-time experiment from trial-level checks to statistics and an author-reviewed draft Results paragraph.
Open demo →Use bundled synthetic fixations to inspect reading-measure calculations and a draft paragraph that still requires author review and study-level modelling.
Open demo →Explore how a synthetic proposal summary can be mapped to published grant criteria for structured pre-submission review.
Open demo →Extract structured fields from 11 pinned abstracts and inspect agreement, omissions and the limits of a small non-exhaustive corpus.
Open demo →Compare glossary-assisted and unassisted translations on 14 synthetic government-style sentences, with terminology hits separated from AI-judged quality.
Open demo →Test how a Cantonese-language prototype routes 15 scripted messages through non-diagnostic distress flags and crisis-handoff rules.
Open demo →Inspect whether an exploratory pipeline recovers the planted three-factor structure, then review its draft Methods and Results wording.
Open demo →Compare two detection methods on a small constructed set and inspect why paired performance does not establish single-text reliability.
Open demo →Re-run the deposited clinical-survey analyses and inspect where an open-source approximation differs from the original invariance test.
Open demo →See how an AI applies a fixed category codebook to 30 public complaint narratives and compare its classifications with the dataset's existing labels.
Open demo →Compare rate, transcript accuracy and pitch variation across two controlled recordings without treating the outputs as child norms or validated assessment
Open demo →The pipeline re-runs the core statistical analyses from a 2022 PLOS ONE study on child-care worker burnout — and checks how closely the numbers match what
Open demo →Re-run the archived interpreting-error analyses and compare the direction and relative importance of effects in a small exploratory sample.
Open demo →Inspect how an AI applies stated inclusion criteria to 16 hand-written abstracts, with every provisional decision kept open to human review.
Open demo →Re-run the archived code and corpus to inspect 20 exact matches and four documented discrepancies.
Open demo →Explore sentiment calibration, theme extraction and summary drafting on a fixed English review corpus, with human review required for transfer.
Open demo →Compare AI judgements across ten controlled translation pairs, including preference, error detection and less-reliable error-type labels.
Open demo →Re-run selected distribution and growth checks on the archived Korean hip-hop dataset and compare the resulting pattern with the published analysis.
Open demo →Inspect turn counts and talk-share estimates from a clean synthetic fixture, without claiming equivalence to a validated research device.
Open demo →Compare selected group-level patterns from an open child-language dataset using an approximate model; this is not a diagnostic or exact software
Open demo →Compare AI scores and feedback with two human ratings on 25 public US middle-school essays, while examining the validation and oversight required before
Open demo →Re-run the deposited analyses and inspect the 20 matched results alongside the one borderline mismatch.
Open demo →Recalculate selected evaluation metrics from shared Shenzhen image ratings and precomputed AI predictions, while flagging an unresolved baseline-version
Open demo →Independently recode 77 public news articles and inspect where automated labels agree—or disagree—with a transparent keyword check.
Open demo →Reanalyse three open behavioural datasets with a population-averaged model and compare directions and significance patterns with the paper's
Open demo →Vary cleaning and analysis decisions in simultaneous EEG and eye-tracking data, then inspect how an underpowered natural-reading result changes.
Open demo →