SurplusLink AI
A prototype B2B marketplace platform for surplus packaged bakery goods targeting the Pakistani market. It uses open-source large models to parse natural-language procurement requirements and combines a rule engine to complete inventory matching.
Timeline (last 90 days)
SurplusLink AI is a prototype B2B trading platform for surplus packaged bakery products targeting the Pakistani market.
SurplusLink AI's initial categories are limited to sealed packaged goods such as biscuits, rusks, crackers and cookies.
SurplusLink AI uses an open-source language model to extract fields such as category, quantity, unit, budget ceiling, city and minimum remaining shelf life from retailers' natural-language requests.
SurplusLink AI uses Pydantic to validate the fields extracted by the language model; missing fields are not guessed and are saved only after user review and confirmation.
SurplusLink AI's backend uses deterministic business rules to perform inventory matching, with matching dimensions including category, unit, quantity, city, shelf life and price.
SurplusLink AI delegates money-related decisions such as pricing and inventory allocation to a testable, explainable rules engine rather than a black-box model.
SurplusLink AI uses PostgreSQL to manage relational data, with SQLAlchemy and Alembic handling model definitions and versioned migrations.
SurplusLink AI defines its inventory availability rule as: available quantity = total quantity - reserved quantity - sold quantity.
When a quote is accepted, SurplusLink AI's backend reserves inventory transactionally to prevent the same batch of goods from being allocated repeatedly to multiple buyers.
SurplusLink AI's AI inference layer calls hosted open-source Llama models through GroqCloud.
4 more timeline events
All Facts (14)
SurplusLink AI is a prototype B2B trading platform for surplus packaged bakery products targeting the Pakistani market.
50%UnverifiedSurplusLink AI's initial categories are limited to sealed packaged goods such as biscuits, rusks, crackers and cookies.
50%UnverifiedSurplusLink AI uses an open-source language model to extract fields such as category, quantity, unit, budget ceiling, city and minimum remaining shelf life from retailers' natural-language requests.
50%UnverifiedSurplusLink AI uses Pydantic to validate the fields extracted by the language model; missing fields are not guessed and are saved only after user review and confirmation.
50%UnverifiedSurplusLink AI's backend uses deterministic business rules to perform inventory matching, with matching dimensions including category, unit, quantity, city, shelf life and price.
50%UnverifiedSurplusLink AI delegates money-related decisions such as pricing and inventory allocation to a testable, explainable rules engine rather than a black-box model.
50%UnverifiedSurplusLink AI uses PostgreSQL to manage relational data, with SQLAlchemy and Alembic handling model definitions and versioned migrations.
50%UnverifiedSurplusLink AI defines its inventory availability rule as: available quantity = total quantity - reserved quantity - sold quantity.
50%UnverifiedWhen a quote is accepted, SurplusLink AI's backend reserves inventory transactionally to prevent the same batch of goods from being allocated repeatedly to multiple buyers.
50%UnverifiedSurplusLink AI's AI inference layer calls hosted open-source Llama models through GroqCloud.
50%UnverifiedSurplusLink AI's matching engine supports bidirectional discovery: buyers can find goods, and sellers can also find demand.
50%UnverifiedSurplusLink AI envisions generating revenue through transaction fees rather than monthly fees: a 2% buyer service fee and a 2% seller commission.
50%UnverifiedTaking a wholesale transaction of 150 packs of biscuits at a unit price of 120 rupees as an example, the transaction subtotal is 18,000 rupees, the buyer and seller each pay a 360-rupee fee, the buyer pays 18,360 rupees in total, the seller receives 17,640 rupees, and the platform's gross revenue is 720 rupees.
50%UnverifiedThe author states that the business scenarios in the article are illustrative examples only, not real revenue, and that the prototype does not process real money transactions.
50%