One platform for food product development and commercialization.

Formulas are tenant-scoped with multiple versions and status lifecycles (draft through published/archived). Line items bind tenant ingredients and quantities; nested components roll up nutrition with cycle protection. Calculated nutrition is stored per version in nutrition snapshots; immutable published-style states can be captured as formula snapshots. Change history and activity feeds surface who changed what, when.

Scenario-style modeling exists in the data model for what-if work; depth of UI around scenarios should be confirmed for your deployment. AI-adjacent formulation runs (structured suggestion and draft flows) appear when the corresponding features are enabled.

• Composition: line items plus nested child formulas or ingredients
• Nutrition rollup from profiles and components into stored snapshots
• Version tags, snapshots, and audit-style activity alongside the record
• Approval workflows integrated with formula version transitions

Process stages (yield, moisture loss, retention) are modeled in the foundation; integration with the live nutrition snapshot pipeline should be treated as partially wired until validated in your environment.

Jump to formulation

Tenant ingredients link into an Ingredient Master library and a global ingredient graph used for search, supplier listings, and tenant vault workflows. Organization-specific nutrient overlays layer on top where you need private data. Substitution suggestions use rules and scoring (function, cost, allergen dimensions) — not a black-box AI rank by default.

Operators configure data sources (USDA FoodData Central, Canadian Nutrient File, FooDB, and others) with background sync jobs, provenance, and governance rules that can skip or protect fields. When embeddings are configured, semantic ingredient discovery augments keyword search, including in Adam-related paths.

• Tenant library + global catalog with aliases and nutrient profiles
• Connectors and sync jobs with import logs and governance
• Rules-based substitution scoring exposed in APIs and formulation UX
• Optional vector search when embeddings are enabled

AI-assisted ranking for substitutions is described internally as a future enhancement; today’s path is rules/metadata-driven scoring.

Jump to ingredient intelligence

Suppliers register, verify email, and (where configured) complete an approval path before using the portal. Inside the portal they maintain company profiles, ingredient listings, experts, users, documents, inquiries, and qualification uploads — with upload limits enforced at the API layer.

Publication uses trust and moderation rules: by default suppliers publish listings when ready; flagged listings or documents are hidden from tenant-facing discovery until resolved. Operators can enable strict mode to queue submissions for review. Tenants discover supplier-published content only when visibility and trust rules allow — this is governed in-product discovery, not open procurement.

• Document taxonomy by type and region; optional embedding for retrieval
• Qualification document flows separate from general ingredient files
• Inquiries and activity surfaced to both sides with scoped access

Jump to supplier portal

Nutrition engine output drives Nutrition Facts–style previews for US (FDA-oriented vertical family) and Canada (bilingual HC/CFIA vertical family, including combined saturated+trans %DV handling). Ingredient statements merge nested sources; allergen buckets cover present, may contain, and facility; claims evaluation uses configured rule sets. Regulatory profiles, label rules, templates, and generated label linkage connect to formula versions. A commercialization-oriented rule engine scores readiness by launch region.

Label output is a technical preview — it is not a certification of compliance. Certain Canadian formats (small package, horizontal, linear, simplified, aggregate), dual-column tables, and front-of-package nutrition symbols are explicitly not generated in the current preview path.

• US and Canada nutrition previews tied to formula versions
• Allergen snapshots and claims evaluation scaffolding
• Regulatory profiles and label artifacts linked to the record

Jump to compliance & labeling

Costing resolves ingredient costs through a documented precedence chain (including supplier costs and fallbacks), normalizes units where the stack expects grams for math, and produces line-level extended costs, totals, cost-per-kg, and contribution percentages. Optional cost snapshots persist named scenarios. Admins can import cost rows via CSV batches; permissions separate viewing from managing cost data.

• Formula version costing APIs and ingredient alternate views
• Snapshot persistence for scenario comparison when enabled
• Role-governed access to sensitive financial fields

Jump to costing

Projects carry stages from concept through commercial, link to an active formula assignment, and include intake with advisory guardrails, team membership, milestones, reports, and decisions that can reference formula versions and insights. Project insights carry categories (regulatory, cost, sensory, and others) with review status; notifications can queue email jobs where configured.

Beyond projects, the tenant workspace includes approval workflows on formula versions, comments with mentions, structured action requests and tasks, and change proposals — so reviews leave a durable trail on the same objects R&D edits.

• Project ↔ formula linkage with intake, milestones, and decisions
• Formula approval steps aligned to version status transitions
• Comments, tasks, and proposals on supported entities

Jump to projects & workflows

Each project can maintain ordered processing instruction rows: free-text steps or ingredient-addition rows tied to formula line items, with optional references to the formula version in force when the sheet was edited. That gives teams a structured handoff sheet for pilot and scale-up conversations seeded from the live BOM.

Adam can consume a server-built summary of those rows for review and Q&A when project context is provided — including optional inference helpers that are explicitly labeled as inference in prompts.

• Ordered steps linked back to formula line items where modeled
• Traceable association to formula versions as specs evolve
• Support for manufacturing prep conversations — not batch release records in an external QMS

Instruction rows are project-scoped in Nuforiq. Whether they replace signed batch documentation is a process decision outside the app.

Nuforiq does not execute or control plant operations — it is not a manufacturing execution system and does not replace an ERP on the shop floor.

Sensory sessions support panel modes, test types, sample order, blinded labels, invites, submissions, and analysis artifacts (metrics, summaries, charts, executive snapshots, strengths and weaknesses, comments). An AI interpretation panel can summarize results where enabled. Sample design locks after certain lifecycle points to preserve data integrity.

• Templates, attributes, and methods modeled for repeatability
• Sessions attach to projects alongside other experiment types
• Results dashboards with exploration tools for panel data

External panelist delivery (email, tokens) depends on deployment configuration and operational hardening.

The stack includes lab template families, analysis runs with status lifecycles, experiment execution workspaces (equipment, weigh states, processing steps, observations, measurements, event logs), and experiment requests (for example pilot-plant or QA support). These structures mirror how pilot and lab work is tracked in code.

• Templates and runs with approval-style states where modeled
• Execution records for equipment, measurements, and observations
• Requests routed with priority and status for cross-team work

Schema coverage is broader than any single screen; validate the exact pages and permissions enabled for your tenant before promising a specific lab workflow end-to-end.

Adam is a natural-language assistant with intent handling, retrieval over structured data and (where enabled) ingested documents, LLM generation, optional tool calls, confidence and warnings, citations, and audit records. Tenant and platform configuration can disable it globally or per capability; provider errors surface as structured retry hints.

Registered tools are whitelisted — for example search and detail for ingredients and suppliers, document retrieval, formula context and analysis, and commercialization scoring — each mapped to permission-aware handlers. A legacy query path deliberately avoids private formula line items and secret tables; full assistant flows use orchestration and tools instead.

• Document ingestion, chunking, and semantic retrieval when enabled
• Tool-gated access aligned to tenant permissions
• Outputs framed as assistance — not autonomous compliance sign-off

Adam is not a regulatory authority: answers depend on models, retrieval coverage, and tool data; confidence is heuristic.

Jump to Adam