A Duviloper product

The AI engine for spectroscopy data.

Point it at any folder. It reads the spectra, converts them to ML-ready CSVs, and gets you to publishable figures in minutes. Built for scientists who would rather do research than learn software.

01
Detect
reads structure, not filenames
→
02
Convert
two linked CSVs, 0% NaN
→
03
Align
every shift recorded
→
04
Preview
click peaks, freeze layers
→
05
Export
paper figures, 300 DPI
Download for Windows See it work
100% local15+ formats6 modalitiesFDA-aligned recordsFree for universities
◈
A model reads the data itself. Spectrum shape, spacing, and axis behavior decide whether a file is Raman, FTIR, or mass spec. File names only get a converter so far; structure makes the call.
01 · Mismatch
~/data/prostate_ftir_ev/  ~/data/mrsa_zenodo/  +32 more
spectra_01.dpt
run7.spc ?
colistinR.npz
ftir.jdx
meta.xlsx
archive.zip
README.txt
maldi.raw ?
plasma.csv
23 file formats across 34 folders. No common schema, no shared axis
convert ~/data → eQC format
✓ prostate_ftir_ev    FTIR    1,204 samples
✓ mrsa_zenodo       Raman    3,480 samples
✓ colistin_kp        MALDI    2,905 samples
83,000 samples · 12 groups · 0% NaN · audit-logged
norm: SNV
❄ s_00087s_00412
(a) s_00412  ─  s_00087 ···
A4 · 300 DPIOkabe–Itofig1.png ✓
One folder in, any format: exports, containers, archives, READMEs. Nothing matches

From Duviloper Core AI Lab. Your spectra and results never leave your machine.

68×
Faster Than Manual Conversion
83,000
Samples Converted in 1h 28m
100 GB
Real Research Data Processed
0
Files Uploaded Anywhere
6
Modalities Auto-Detected
21 CFR
Part 11 Record Design
Backed by Inception
See The Difference

Same data. Two paths.

The right side runs unattended: format detection, modality routing, axis fixes, metadata recovery. Same 34 folders, two routes.

Manual 00:00
Folder progress: 0 / 34
Click "Run Comparison" to start.
eQCX 00:00
Folder progress: 0 / 34  |  Samples: 0
Click "Run Comparison" to start.
Supported Formats

If it contains spectra, we handle it

Text exports, industry standards, binary containers. Unknown layouts are handled by structure, not a format table.

.csv
Wide + narrow tables
.tsv / .tab
Tab-delimited
.dpt / .asc
Bruker text export
.jdx / .dx
JCAMP-DX (AFFN + SQZ)
.spc
Galactic/Thermo (3-tier)
.npz / .npy
NumPy arrays
.xlsx
Excel spreadsheets
.rds / .rda
R data files
.mat
MATLAB matrices
.pkl
Python pickle
.zip
Auto-extracted
.tar.gz
Auto-extracted
+ more
Unknown formats handled structurally
The Engine

Raw exports in. Paper figures out.

Every step below is a click in the app, not a manual.

1 · It reads your data
Goal-directed discoveryFinds the axis, intensities, and sample IDs. Uncatalogued formats still convert.
Modality detectionRaman, FTIR, NIR, MALDI, SERS, UVRR. Decided by the data, not filenames.
Hidden metadata miningReads READMEs and headers for facts other converters skip.
Per-sample metadataEvery descriptive column lands on the right sample row.
Tiny model, any laptopAbout 500 parameters. No GPU. Runs anywhere.
2 · It prepares it
100% data loyaltyNo compression, no dropped points, 0% NaN in range.
Smart dedupSHA-256 catches duplicate files and zip double-stores.
Mixed ranges, no NaNDifferent ranges split automatically. One file instead? Unified with a fidelity guard.PRO
Scientific correctionsMissing calibration gets reconstructed. The assumption gets logged.
FDA-aligned recordsHash-chained audit trail. Designed for 21 CFR Part 11.
Parallel runsFive workers. Big sets auto-chunked with zero loss.PRO
3 · It shows and ships
Interactive previewThousands of spectra, smooth list, keyboard-first navigation.
Peak selection + frozen layersClick to mark peaks. Pin samples in color and keep browsing.
Scientific axesFTIR descends 4000→400 cm⁻¹. Raman ascends. m/z with base-peak %.
Waterfall figuresStacked FTIR and Raman sets with per-spectrum normalization.
Paper export300 DPI, colorblind-safe palette, caption built in.
4 · It stays yours
100% localSpectra never upload. Enterprise builds run fully offline.
Open format, no lock-inTwo documented CSVs plus an open SDK to verify them.
ReproducibleByte-identical output. Checksums you can re-check independently.
Opens in ExcelTwo plain CSVs. Every number stays a number.
Audit Trail & Scientific Corrections

Every conversion is a record. Every correction is logged.

Missing data gets a stated assumption, and the assumption gets logged. A real MALDI-TOF conversion, before and after:

BEFORE: Source file (no m/z axis found)
Source: colistinR_plot213_20220426.npz Keys: arr_0, arr_1, ... arr_201 (202 arrays) Shape: each (18000,) float64 Values: 0.0 to 1.0 (normalized) Axis array: NOT FOUND m/z calibration: MISSING from npz # The npz contains 202 spectra but NO axis array. # Without calibration, m/z values are unknown. # How does the model know these are MALDI?
→
AFTER: eQC output with scientific inference + logged correction
Software: eQCX v1 Operator: Dr. Smith Source: colistinR_plot213_20220426.npz Detected: MALDI-TOF MS (1853 samples, 18000 pts) ⚠ SCIENTIFIC CORRECTION APPLIED Issue: No m/z axis array in source npz Inference: Values normalized 0–1 (TIC-normalized MALDI export) Point count ≥ 1000 (Bruker Biotyper standard) Correction: Applied Bruker MALDI calibration m/z = linspace(2000, 20000, 18000) Da Confidence: HIGH. Matches Bruker Biotyper default acquisition Axis calibration assumptions: assumed_brucker_maldi_2000_20000: No m/z axis array found. Applied the industry-standard Bruker MALDI Biotyper calibration (2000–20000 Da). Basis: (1) normalized 0–1 values; (2) point count ≥1000; (3) no monotonic grid-like axis present. Verify against original instrument calibration if precision is critical. Output: m_z group, 1853 samples, 0.0000 NaN SHA-256: 7a3e9f1c...e2b8 Status: SUCCESS

What corrections can the model make?

m/z Calibration Recovery

No axis in the file? The Bruker Biotyper standard is applied, and the assumption logged.

µm → cm⁻¹ Conversion

JCAMP micrometers converted to wavenumbers. Logged.

Sample ID Generation

Best descriptive column wins; otherwise one is generated. Decision logged.

Axis-Range Separation

Mixed spectral ranges split into separate files. No NaN padding.

Modality Classification

Decided from the spectrum. Ambiguous cases state their default.

Transposed Container Fix

Sideways arrays flipped automatically. Correction logged.

Pricing

Start free. Scale when you're ready.

The free tier needs no card. Pay when you need unified grids, alignment, or batch scale.

Researcher
Students, academics, individual researchers
Free
  • ✓ Single-folder conversion
  • ✓ All 15 supported formats
  • ✓ Metadata enrichment + per-sample features
  • ✓ FDA-aligned compliance sheet
  • ✓ Desktop app (Windows)
  • ✗ Unification + alignment
  • ✗ Commercial use
Download Free
Lab
Research labs, small teams, CROs
$15/month
  • ✓ Everything in Researcher
  • ✓ Wavenumber unification to one file
  • ✓ Peak alignment, shifts recorded
  • ✓ Unlimited batch
  • ✓ Full audit history
  • ✓ 5 parallel workers
  • ✓ Commercial license
  • ✓ Email support (48h)
Enterprise
Pharma, biotech, clinical labs
$500/site/year
  • ✓ Everything in Lab
  • ✓ 50 devices per site
  • ✓ On-premise, fully isolated
  • ✓ Data anonymization (coming)
  • ✓ Pipeline API (coming)
  • ✓ Custom format adapters
  • ✓ Part 11-designed audit records
  • ✓ BigQuery / SQL ingestion (coming)
  • ✓ Dedicated support (24h)
Education

Free limited seats for universities

Universities get 20 free seats a month. Sign in to view your license and remaining credits, or check your domain below for open seats.

Sign in

Don’t have an account? Check your university’s seats below.

Each university is allocated 20 seats per month; the allocation resets at the start of each month. When the allocation is exhausted, further registrations are placed on a waiting list and are admitted when seats reset or the university adopts a campus plan.

Runs locally. Your spectral data and conversion results never leave your machine. Only your email, a pseudonymous usage ID, and quota counts travel. Passwords are stored as salted PBKDF2 hashes and never in readable form.

Services on Demand

Need more than a tool?
We build the infrastructure.

For organizations that need more than software.

Dedicated Cloud Deployment

Your cloud account, GPU instances, managed pipeline.

From $2,000/month · Custom quote
Big Data Pipeline Engineering

eQC output into BigQuery, Snowflake, or your data lake.

From $5,000 · One-time + maintenance
Full ML Infrastructure

Conversion to trained model, end to end.

Custom quote · Tailored to your needs
Custom Format Adapters

Proprietary formats reverse-engineered, built, and maintained.

From $5,000 · Per format
FDA Validation Package

IQ/OQ/PQ documents and Part 11 evidence your QA team needs.

From $15,000 · Per site
Training & Onboarding

Half-day workshops: the app, the audit trail, ML on eQC output.

From $2,500 · Half-day session
Discuss Your Needs →

Every engagement is scoped to your instruments, data volume, and infrastructure. No two labs have the same setup. We adapt the solution to fit.