Every allergen, accounted for — from supplier spec to printed label.
Allersift reads your supplier specifications, finds allergens hidden in ingredient lists, and checks them against your recipes and packaging labels — so your QA team reviews flagged evidence instead of reading every page.
Ingredients: WHEAT flour, sugar, cocoa compound coating (sugar, palm fat, cocoa, SOYA lecithin, skimmed milk powder), vegetable oil, raising agents.
Contains: wheat, soya.
Milk detected — missing from "Contains"
Source: supplier spec SPEC-2042 (cocoa coating), page 2. Sent to QA review.
01 — The problem
Allergen errors start in documents nobody had time to read twice.
A mid-sized food manufacturer handles hundreds of supplier specifications, each a different format, many scanned. Allergen information sits in ingredient lists, declaration tables and footnotes — and they don't always agree.
QA teams copy this into spreadsheets by hand. When a supplier swaps a coating that now contains milk powder, the change has to be noticed, traced to every recipe that uses it, and caught before label artwork goes to print.
Missed allergens lead to product recalls, retailer delistings, wasted packaging and, most seriously, harm to consumers with allergies. Manual checking doesn't grow with SKU count.
02 — The platform
Six capabilities, one allergen record per product.
Each capability feeds the next: documents become structured ingredients, ingredients become allergen findings, findings are reviewed by people, and reviewed results become the record you show auditors.
- 01
Supplier document reading
Upload specifications, certificates of analysis and allergen questionnaires in PDF or scanned form. Ingredient lists, allergen tables and cross-contact statements are extracted into structured fields.
No more retyping spec sheets into spreadsheets.
- 02
Allergen risk detection
Each ingredient is checked against the major allergen groups, including derived names such as casein, lecithin, kuli-kuli or semolina, and flagged as contains, may contain or not detected.
Hidden and renamed allergens surface before production.
- 03
Formulation-to-label cross-check
Recipes are linked to the raw materials they use. When a supplier changes a material, every affected product and label is checked again.
A coating change no longer quietly breaks a "Contains" statement.
- 04
Human review queue
Every AI finding is routed to a named reviewer with the source text, page reference and reasoning. Nothing is approved automatically.
QA stays in control; the software does the reading.
- 05
Supplier document tracking
See which suppliers have outdated specifications or missing declarations, and send requests for updated documents.
Supplier files stay current between audits.
- 06
Compliance reports
Export a product's allergen record, with source documents, reviewer decisions and dates, for audits and customer requests.
Audit preparation in minutes instead of days.
03 — How it works
From a supplier PDF to a signed-off allergen record.
STEP 1
Submit
Supplier documents, recipes and label artwork are uploaded or emailed in.
STEP 2
Extract
Text, tables and statements are pulled from every page, including scans.
STEP 3
Analyse
AI maps ingredients to allergen groups and compares declarations, recipes and labels.
STEP 4
Review
Your QA team confirms or rejects each finding with the evidence beside it.
STEP 5
Record
Approved results become a dated compliance record linked to each product.
Allersift supports human decision-making. It does not replace your HACCP plan, food safety team or regulatory advice, and final labelling decisions stay with your qualified staff.
04 — Interactive preview
Try the compliance workspace.
Search documents, filter by risk, open a record to see highlighted allergens and the AI explanation, then approve or return it. This preview runs on fictional sample data in your browser.
| Document | Supplier / item | Detected | Risk | Status |
|---|---|---|---|---|
| SPEC-2041 Specification | Fortified wheat flour (Type 550) Harmattan Mills Ltd | Wheat/Gluten | Low | Approved |
| SPEC-2042 Specification | Cocoa compound coating Lagoon Cocoa Processors | SoyMilk | High | Needs review |
| COA-0918 Certificate of Analysis | Suya spice blend Eko Spice Blenders | Peanut | High | Needs review |
| SPEC-2039 Specification | Sweetened condensed milk Northbank Dairies | Milk | Medium | In progress |
| LBL-0311 Label artwork | Choco Crunch Biscuit 150g — label proof v3 In-house | Wheat/GlutenSoyMilk | High | Needs review |
| SPEC-2036 Specification | Refined groundnut oil Delta Oils Co. | Peanut | Medium | Approved |
| SPEC-2033 Specification | Pasteurised liquid whole egg Ibadan Egg Products | Egg | Low | Approved |
| SPEC-2030 Specification | Shrimp seasoning powder Crestwave Seafoods | CrustaceansWheat/Gluten | Medium | In progress |
05 — Who it's for
Built for the people who sign off on allergen labels.
Packaged food manufacturers
Challenge: Dozens of SKUs, hundreds of raw materials, frequent supplier substitutions.
With Allersift: Link every recipe to its supplier specs and re-check labels whenever a material changes.
Bakeries and confectioners
Challenge: Shared lines with wheat, milk, egg, nuts and sesame; precautionary labelling decisions.
With Allersift: Track cross-contact statements per material and keep "may contain" wording consistent.
Quality and regulatory teams
Challenge: Audit requests, customer questionnaires and spec reviews done by hand.
With Allersift: Use the review queue as the single place where allergen decisions are made and recorded.
Private-label and contract producers
Challenge: Producing to retailer specifications that demand documented allergen control.
With Allersift: Export product-level allergen reports with source evidence for each retailer.
06 — Technology
How We Use Amazon Web Services
Allersift's MVP architecture is designed on AWS around one path: a supplier document arrives, its text is extracted, allergens are interpreted, and the result is stored as a reviewable compliance record. The serverless design keeps costs tied to document volume while the product is being built.
- Amazon S3: Stores uploaded supplier specifications, certificates and label artwork with encryption at rest.
- Amazon Textract: Extracts ingredient lists, allergen tables and form fields from PDFs and scanned documents.
- AWS Lambda: Runs the processing steps: triggering extraction, normalising ingredients and comparing recipes against labels.
- Amazon Bedrock: Interprets extracted text to map ingredients to allergen groups, read cross-contact statements and write reviewer-facing explanations.
- Amazon DynamoDB: Holds structured compliance records: materials, recipes, findings, reviewer decisions and audit history.
07 — Progress & early access
We're building the MVP with food manufacturers.
- 2026Allersift founded
- NowProduct design, interactive preview and MVP architecture
- NextDocument extraction and allergen detection with early-access manufacturers
Early-access partners help shape which document types and allergen rules we support first. Tell us about your products and suppliers.
08 — Pricing
Simple monthly plans in naira.
Starter
₦5,000 / month
For a single site getting supplier documents organised.
- Up to 50 supplier documents a month
- Allergen detection and flags
- Review queue for 2 reviewers
- Product allergen summaries
Growth
Most suited to SMEs₦10,000 / month
For manufacturers managing many recipes and suppliers.
- Up to 250 supplier documents a month
- Recipe-to-label cross-checks
- Supplier document tracking
- Unlimited reviewers
- Audit-ready report exports
Enterprise
₦20,000 / month
For multi-site producers and contract manufacturers.
- Higher document volumes
- Multiple sites and product lines
- Custom allergen rule sets
- Priority onboarding support
Plans become available as the MVP opens to early-access customers.
09 — Team
The Team Behind Allersift
Abass Abdulazeez
Founder
Abass founded Allersift in 2026 to give food manufacturers a faster, more reliable way to verify allergen information across supplier documents, recipes and labels. He leads product, engineering and early customer conversations.