POC Results

RNDA domain-specific genomic encoder. Highest compression of any data type tested. Species-level discrimination proven — same species similarity ~1.000, cross-species ~0.04. Raw FASTQ files permanently discarded after encoding.

Source: NCBI public sequences — 28 files across 6 species (bacteria, human, mammal, plant, fungi, insect)

RNDA domain-specific genomic encoder. Same organism similarity 1.0000. Different organism 0.06. Discrimination gap 0.939. Proven on 300 real NCBI RefSeq bacterial genomes.

Source: NCBI RefSeq bacterial genomes — 300 real sequences

RNDA domain-specific point cloud encoder. 3,356 real Velodyne frames across city, residential, and road environments. Same-drive clustering perfect. Raw sensor data permanently discarded.

Source: KITTI autonomous driving dataset — 9 sequences across 4 dates

RNDA domain-specific financial encoder. Binary market pattern matching proven — discrimination gap 0.99. Submit 30-day return pattern, find most similar historical conditions. Market regime detection, M&A target analysis, regulatory surveillance.

Source: Real OHLCV — 50 S&P 500 tickers, 5 years

RNDA domain-specific satellite encoder. Binary scene matching proven — discrimination gap 0.98. Forest finds Forest, Residential clusters correctly. Submit satellite image, find similar historical scenes.

Source: Real satellite imagery — 10 terrain and land use types

RNDA domain-specific multi-sensor encoder. 6 cameras (360°) + LiDAR + 5 radar units fused into single signature. Best discrimination gap of all domains tested — 1.09. Raw sensor data permanently discarded.

Source: Real synchronized AV sensor data — 6 cameras + LiDAR + 5 radar units fused

2,000 real NIH chest X-rays. Same disease surfaces at top of results every time. Discrimination gap 0.82. 1,235/2,000 images score <0.3 (clearly different). RNDA domain-specific medical imaging encoder.

Source: NIH ChestX-ray14 dataset — real labeled chest X-rays

RNDA domain-specific audio encoder. Real FLAC audio recordings. Binary audio queries proven — discrimination gap 0.95. Submit audio, find similar recordings. Raw audio permanently discarded.

Source: Real FLAC audio files

RNDA domain-specific video encoder. 175 real action video clips across 35 categories. Binary video queries proven — discrimination gap 0.998. Same-action clips cluster perfectly. Raw video permanently discarded after encoding. Applications: content fingerprinting, surveillance archiving, sports analytics, streaming platform IP protection.

Source: Real video clips — 35 action categories, UCF-101 public dataset

Real seismic waveforms from public IRIS database. Binary waveform queries proven — discrimination gap 0.97. Same geological station: 0.4–1.0. Different stations: near zero. Submit waveform, find similar formations.

Source: Real IRIS/EarthScope waveform data

9,379 real 3D molecular structures. RNDA domain-specific molecular encoder. Structural similarity search.

Source: Real 3D molecular structures (SDF format)

RNDA domain-specific code encoder. 576 real source files across 3 languages — Python, JavaScript, Java. Binary code similarity proven — discrimination gap 1.023. Same-language files cluster perfectly. Applications: code plagiarism detection, IP protection, software compliance, license auditing.

Source: Real open source repositories — Python, JavaScript, Java

Real financial and corporate communication networks. Binary network pattern matching proven — discrimination gap 0.98. Find similar network patterns for fraud detection, M&A coordination analysis, regulatory surveillance.

Source: Enron email network (367K edges) + Bitcoin trust network (24K transactions)

RNDA domain-specific timeseries encoder. Real industrial sensor data — accelerometer/gyroscope readings. Binary queries proven — discrimination gap 0.99. Same activity clusters at 0.7–1.0 similarity, different activities near zero.

Source: Real sensor and industrial time series data

RNDA domain-specific industrial encoder. CNC milling, water pump, semiconductor manufacturing sensors. Binary pattern matching — discrimination gap 1.00. Equipment type clustering perfect.

Source: Real industrial sensor data — 3 manufacturing systems

RNDA domain-specific multi-sensor encoder. 6 cameras (360°) + LiDAR + 5 radar units fused. Best discrimination gap of all domains tested — 1.09.

Source: Real synchronized autonomous vehicle sensor data

RNDA domain-specific brain signal encoder. Raw EDF clinical files averaging 1.9MB each. Motor imagery vs rest discrimination proven.

Source: Real clinical EEG recordings — motor imagery and rest tasks

RNDA domain-specific MRI encoder. Gray matter / white matter / fMRI discrimination proven. Raw NIfTI files averaging 1.1MB each.

Source: Real T1-weighted and fMRI brain scans — 405 real clinical scans

RNDA domain-specific spectral encoder. Raw FITS spectral files. Galaxy, star, and quasar discrimination proven.

Source: Real spectroscopic observations — galaxies, stars, quasars

RNDA domain-specific structural encoder. Raw PDB files with full 3D atomic coordinates averaging 1MB each.

Source: Real 3D protein coordinate files from public research database

RNDA domain-specific climate encoder. Arctic, desert, tropical, temperate zones. Arctic 8/10, temperate 10/10 same-zone clustering.

Source: Real weather station data — 76 global locations, full year

RNDA domain-specific temporal graph encoder. Network type discrimination perfect. Applications: fraud detection, compliance surveillance.

Source: Real corporate email network (367K edges) + financial trust network

RNDA domain-specific legal encoder. Full 10-K annual reports at 911x compression (700KB+ files). Company-specific clustering proven.

Source: Real SEC 10-K annual filings — major public companies

RNDA domain-specific signal encoder. 480,000 real IQ samples across 7 modulation types. Binary signal queries proven — discrimination gap 1.009. Same modulation type clusters perfectly. Submit signal, find similar transmissions. Applications: spectrum monitoring, signal intelligence, communications analysis.

Source: Real IQ signal recordings — 7 modulation types, multiple SNR levels

RNDA domain-specific IoT encoder. Real continuous sensor streams from industrial systems. Binary pattern matching proven — discrimination gap 1.027. Sensor type discrimination perfect. Applications: anomaly detection, predictive maintenance, supply chain monitoring.

Source: Real industrial sensor streams — temperature, traffic, CPU load monitoring

RNDA domain-specific clinical encoder. 200 real ICU patients — continuous HR, BP, SpO2, temperature streams. Binary patient similarity proven — discrimination gap 0.762. Raw patient data permanently discarded after encoding. Applications: clinical pattern matching, critical care surveillance, patient cohort analysis.

Source: Real ICU patient vital sign streams — 200 patients from public clinical database

RNDA domain-specific retail encoder. 541,910 real transactions from 4,338 customers across 38 countries. Binary customer behavior fingerprinting proven — discrimination gap 0.947. Find similar customer segments without retaining any purchase history. Applications: personalization at scale, loyalty program analytics, fraud detection, GDPR-compliant customer intelligence.

Source: Real retail transaction data — 541,910 purchases across 38 countries

RNDA domain-specific network encoder. Real uncompressed PCAP files — HTTP, HTTPS, DNS, SSH traffic. Binary traffic fingerprinting proven — discrimination gap 1.049. Same traffic type clusters perfectly. 5MB raw captures → 256-byte signatures. Enterprise POC runs on client network at scale. Applications: carrier traffic analysis, network security, compliance logging, traffic classification without raw packet retention.

Source: Real uncompressed network packet captures — 4 traffic types, live production server

RNDA domain-specific telecom encoder. 7,043 real subscriber profiles across phone, internet, and streaming services. Binary subscriber pattern matching proven — discrimination gap 0.970. Identify similar customer segments without storing raw subscriber records. Applications: churn prediction, network capacity planning, subscriber segmentation, CCPA/GDPR compliance.

Source: Real telecom subscriber data — 7,043 customers, 21 service and usage dimensions

Peak compression on 90MB large-scale circuit. Dataset includes circuits verified on IBM Burlington, Melbourne, and Paris quantum backends.

Source: QASMBench (PNNL) — verified on IBM Quantum hardware

RNDA domain-specific well log encoder. 118 real Norwegian North Sea well logs averaging 5.2MB each. Binary well similarity proven — discrimination gap 0.731. 20,448x compression — multi-MB raw logs to 256-byte signatures. Applications: formation evaluation, reservoir characterization, well-to-well similarity search.

Source: Real LAS well log files — Norwegian North Sea, public research dataset