CASE FILE • AI-POWERED ML FORENSICS

Find the failure.
Trace the evidence.

ReproLens investigates your ML experiment like a detective investigates a case — connecting artifacts, exposing contradictions, and reconstructing the most likely root cause.

CASE #RL-2048•UNDER INVESTIGATION
Explore How It Works
config.yaml
dataset_metadata.json
train.log
requirements.txt
traceback.txt
ReproLens AI
Root Cause
Evidence

Turn a failed run into a solvable case.

ReproLens doesn't just read a traceback. It reconstructs the chain of events across your experiment and separates verified clues from possible explanations.

01
THE CRIME SCENE

Collect the clues.

Config, dataset metadata, logs, dependencies and tracebacks become one investigation file.

config.yamltrain.logtraceback.txt
02
CONNECT THE DOTS

Find contradictions.

ReproLens compares signals across artifacts instead of treating every file as an isolated clue.

Cross-artifact correlation
03
RECONSTRUCT THE EVENT

Trace the root cause.

Verified facts, competing hypotheses and AI reasoning are assembled into a defensible finding.

Evidence-backed diagnosis

See exactly why the case was solved.

Every important conclusion points back to a source. The interface makes the reasoning inspectable instead of hiding it behind an AI response.

dataset_metadata.json 12 classes detected
train.log training interrupted
config.yamlCLUE
model:
  name: resnet50
  num_classes: 10
  pretrained: true
VERIFIED CONNECTION

Configuration mismatch detected.

The model expects 10 classes while the dataset metadata reports 12. Two independent artifacts point to the same contradiction.

config.yaml10
→
dataset_metadata.json12

Not just an answer. A case you can defend.

When an experiment fails, knowing the error is not enough. You need to know what supports the conclusion and what to check next.

A

Verified first

Deterministic preflight checks establish concrete facts before AI reasoning begins.

B

Correlated evidence

Hidden contradictions emerge when configs, data, runtime signals and logs are examined together.

C

Explainable verdict

Every primary finding comes with evidence, confidence, alternatives and recommended next checks.

Investigations worth reopening.

ARCHIVE

Previous experiments stay searchable as case files — ready to inspect, compare and explain again.

New Investigation

Upload the artifacts that describe your experiment.

Experiment evidence 0 of 5 uploaded

At least one artifact is required. Missing artifacts are surfaced as missing information — they don't fail the investigation.

Investigating…

Evidence bundle
Preflight checks
AI Investigation
Ingesting artifacts
Waiting…
Correlating evidence
Deterministic checks
AI investigation
Bedrock reasoning

Each stage only updates after the backend completes it. Nothing is simulated.

Investigation

Status
Confidence
Artifacts
Findings
Primary Finding

Pre-investigation Checks

Evidence Graph

Evidence Explorer

Artifacts

              

Select an artifact or click an evidence node to inspect source.

Source
—
Location
—
Significance
—

Alternative Explanations

Recommended Next Checks

Ask about this investigation

Investigation History

Review past investigations and their findings.

Example Cases

Pre-loaded fixture files that run through the real investigation pipeline. Nothing is canned.

Silent generalization failure

Training completes with no exception, but validation performance quietly falls apart.

Dataset / config mismatch

The model is configured for one class count; the dataset reports another.

CUDA incompatibility

The runtime CUDA version doesn't match the compiled extension's target version.

Contradictory evidence

Three different class counts across three artifacts — no single confident guess.

Healthy experiment

Internally consistent artifacts — proves ReproLens doesn't manufacture problems.

Investigation Interrupted

Investigation ID: