
TECHNOLOGIES
The software, standards and research behind Digital Food 2.0.
Organised along the chain from recipe to consumption, with two supporting dimensions: the intelligence layer that connects the stages, and the living systems where ideas are tested.
● AVAILABLE Public tools, specifications or services you can use now. Usable does not mean certified or validated. ● EXPERIMENTAL Working prototypes and proofs of concept, not commercial products. ● RESEARCH Research directions and historical frameworks, published as they mature. ○ FUTURE Stated plans, not yet built.
ARCHITECTURE
01 · Recipe
JRF · RecipeAnalyzer 2.0 · TT-o
02 · Composition
Ajinoverse 4.4 · TasteTuner · CACT · FlavorGraph
03 · Experience
MP6 standard & tools · FEELLLM
04 · Consumption
WillULikeIT · NutriSherpa · AayuSense
Intelligence & infrastructure
SensoryOS · AI & sensory intelligence · SRS-KnowledgeGraphDB · weak-signal research · AJX Studio
Living systems & experimentation
AJX BioSphere · BioSensei · environmental sensing and automation
STAGE 01
Recipe
How a dish or product is made — ingredients, quantities, operations, temperatures, timings and intermediate states — captured as data a machine can work with.
JRF — JSON Recipe Format
● AVAILABLE
Turns a recipe into a structured process rather than a paragraph of instructions. A .jrf file is designed to stay meaningful on its own, independent of any Ajinomatrix server. Documented in the JRF/ORS working paper series (v0.1.2).
RecipeAnalyzer 2.0
● AVAILABLE
Estimates how ingredient and process changes shift aroma and texture, so formulation questions can be explored as reasoned “what-ifs”. Built on digital-gastronomy IP licensed from the Hebrew University of Jerusalem. Live online in its legacy form, and integrated in Ajinoverse 4.4.
TT-o — ingredient origin
● EXPERIMENTAL
Carries ingredient origin, supplier, lot and process history alongside a recipe — first applied to cocoa sourcing.
STAGE 02
Composition
What a product contains, how formulation alternatives compare, and how a target profile can be approached without losing the product’s identity.
Ajinoverse
● AVAILABLE
Ajinomatrix’s software suite. Version 4.4, the latest release, integrates RecipeAnalyzer, TasteTuner and the other applications in one environment for producing, inspecting, validating and transforming JRF and MP6 objects.
ajinoverse.org →
TasteTuner
● AVAILABLE
Adapts a recipe step by step toward a target sensory profile while preserving its identity, and visualises the result as a box chart. First developed in an EIT Digital project with the University of Helsinki. Live online in its legacy form, and integrated in Ajinoverse 4.4.
CACT — Chemo-Analytic Clustering Tool
● EXPERIMENTAL
Clusters chemical and analytical data to support sensory prediction — first for aromas, now extended to liquids and solids.
FlavorGraph integration
● RESEARCH
In development: brings the published FlavorGraph approach — a large-scale food–chemical graph for ingredient pairing (Park et al., Korea University and Sony AI, Scientific Reports, 2021) — into the Ajinomatrix stack.
STAGE 03
Experience
How a product is perceived — across sight, smell, taste, touch, sound and intuition — recorded, exchanged, replayed and compared.
THE STANDARD
MP6 — Multi-Perception 6
● AVAILABLE
An open file format for sensory records. The v0.1 public specification (2026) is a compatibility layer: it encourages interoperability, while Ajinomatrix’s internal mapping and decision logic stay protected. A published specification, not a certification.
THE TOOLS · MP6 STUDIO
● AVAILABLE
Converter — Excel, CSV and sensory datasets to MP6.
Recorder — Spoken or typed impressions to a dynamic MP6; Expected MP6 from a JRF recipe.
Viewer — Opens and inspects MP6 files.
Validator — Checks structural validity and compatibility.
Timeline Player — Plays a sensory experience over time.
Compare — Puts two MP6 profiles side by side.
Report Generator — Readable, shareable reports from MP6 files.
Portal — Entry point to the Studio at mp6.studio.
Free public tools from an experimental ecosystem. MP6 Recorder: concept inspired by Markus Ojala, SeAMK University.
AN APPLICATION
FEELLLM
● EXPERIMENTAL
A self-report surface that turns a tasting experience into a structured “Feeling Brief” — context a general AI assistant can use, without claiming the AI itself perceives.
STAGE 04
Consumption
Who eats the result: personal fit, adaptation and longer-term patterns. Proofs of concept only — no clinical claims.
WillULikeIT
● EXPERIMENTAL
Compares a predicted sensory experience with the preferences you state, and returns a transparent Personal Sensory Fit — an inspectable estimate, not a probability. Then you taste it, and reality gets a vote.
NutriSherpa
● EXPERIMENTAL
Proof of concept for kidney-conscious recipe adaptation. Not a medical device and not a substitute for clinical advice.
AayuSense
● EXPERIMENTAL
Proof of concept that looks at where a person’s food pattern is heading, rather than counting calories.
CROSS-CUTTING DIMENSION · INTELLIGENCE & INFRASTRUCTURE
SensoryOS — the sensory infrastructure layer.
A modular, Linux-based operating layer designed to let devices, sensors and AI systems capture, store and query taste, smell, texture and other sensory signals with the rigour already applied to images and audio.
● EXPERIMENTAL
Experimental · proof of concept · Not a public product
DEVELOPMENT HISTORY
2023 — Concept introduced; framework first wired out alongside TasteTuner in the EIT Digital project.
Jan 2025 — Architecture published: Linux-based and modular, with a low-power branch for Raspberry Pi-class devices and a high-compute branch.
Sep 2025 — Proof of concept: first autonomous boot on an Ubuntu 22.04 core at the AJXHQ Lab (theEgg, Brussels), processing sample data from BioSphere systems.
Planned — More device integrations and data connectors for sensory panels and GC-MS pipelines; public modules under a dual licence.
HOW IT RELATES
Sensory modelling — Provides the capture-and-store primitives that sensory models need as input.
TasteTuner, RecipeAnalyzer, FEELLLM — Described as the application branches that run on top of SensoryOS.
MP6 — The file format SensoryOS uses to store and exchange sensory records.
Digital Food 2.0 — The intended device and data layer beneath all four stages and BioSphere.
STATUS
Implemented — Bootable proof-of-concept image; sample BioSphere feeds.
Historical — 2023 concept and Raspberry Pi mock-ups.
Experimental — Live sensor feeds, AI integration, community modules.
Proposed — Industrial panel and GC-MS connectors; device ecosystem.
AI & SENSORY INTELLIGENCE
What the computation actually does — and what it doesn’t.
Predictions are frozen before tasting, then compared with what people actually perceive. Errors and rejected suggestions are kept as evidence, and a dimension that wasn’t observed is never silently scored as zero.
Sensory data interpretation
Structures panel sheets, e-nose, GC-MS and texture data; separates what is observed, what can be inferred and which test to run next.
Decision Reports, MP6 Converter · IMPLEMENTED
Recipe & formulation analysis
Estimates how ingredient and process changes shift aroma and texture; steers a recipe toward a target profile.
RecipeAnalyzer, TasteTuner, CACT · IMPLEMENTED
Predictive sensory modelling
Generates an Expected MP6 from a JRF recipe using a published deterministic baseline; a human reviews it before it is frozen.
MP6 Recorder, JRF · EXPERIMENTAL
Comparative evaluation
Compares expected with observed, or one version with another, dimension by dimension.
MP6 Compare, Report · IMPLEMENTED
Knowledge graphs
A provenance-aware graph linking ingredients, compounds, recipes, sensory records and their sources; ingredient-pairing representations. Release planned for the end of 2026.
SRS-KnowledgeGraphDB, FlavorGraph · RESEARCH
Intelligent workflow assistance
Packages context so a general AI assistant can help, and guides users through recipe-to-tasting steps.
FEELLLM, AJX Studio · EXPERIMENTAL
Personalised food
Matches predicted experiences to stated preferences; explores adaptation and food-pattern trajectories. No clinical validation.
WillULikeIT, NutriSherpa, AayuSense · EXPERIMENTAL
Research & decision support
Looks at the edges of datasets — outliers, unusual responses, early weak signals — rather than only their averages.
ABED-SP’G, Black Swan PreCog · RESEARCH
MORE OF THE INTELLIGENCE LAYER
Services, research and what comes next.
AJX Studio
● EXPERIMENTAL
A guided Recipe & Kitchen Lab workflow across all four stages.
Decision Reports
● AVAILABLE
A service: Analytical Gap Scans that turn existing lab and panel data into the next test to run.
SRS-KnowledgeGraphDB
● RESEARCH
A genomic-grade knowledge-graph database for food and sensory data, with provenance. In development; release planned for the end of 2026.
Predictive cuisine
● EXPERIMENTAL
Recipe → Expected MP6 → cooking → Observed MP6 → comparison, live as an open workflow.
Black Swan PreCog
● RESEARCH
Spin-off research on weak signals that precede disruptions such as food crises.
Integrated suite
● FUTURE
One Digital Food 2.0 environment connecting all four stages end to end.
CROSS-CUTTING DIMENSION · LIVING SYSTEMS & EXPERIMENTATION
Where the stack meets living systems.
AJX BioSphere
● RESEARCH
A living testbed for small autonomous food ecosystems. In 2025, a solar-powered installation ran a 61-day autonomous loop from device to growth to dish.
BioSensei
● EXPERIMENTAL
An operating stack for living systems — observe, log, interpret, act — with a SysOps app for monitoring solar fountains, aquaria and greenhouses.
BioSphere syllabus
● AVAILABLE
A low-cost, modular curriculum for learning automation and systems thinking with commodity components, published as a book.
Not sure which tool fits your project?
Sensory Project Fit / MP6 Consultation · 15 minutes · €90, credited against a first Ajinomatrix Sprint.