AI for tenders, employer's product
Tender review with AI
An employer's product for tender specialists: AI reads the whole procurement and says whether it is worth bidding.
- 78
- seen
- 51
- reviewed
- 22
- shortlisted
- ProblemA tender specialist spends hours reading documents only to learn the tender does not fit. The goal was to filter that out before a human starts reading.
- What I did
- Document analysis and go/no-go scoring against a rubric, with reasons.
- A shortlist for the team: only tenders that pass scoring go to work.
- Several AI providers switchable without a rebuild, and two-stage protection against instructions injected into documents.
- Go backend on Clean Architecture and DDD, React and TypeScript frontend, tests gate merges in CI.
- ResultRun on Oct 13: 78 tenders viewed, 51 analyzed, 22 shortlisted. Tender specialists work from the shortlist instead of the full stream.
- Stack and approachGo, React, TypeScript, Clean Architecture, DDD, multiple LLM providers, CI with race tests and linter.
Discuss a similar project
Describe what hurts and I will reply on Telegram with a first step.