tenderMania

Inspiration

TenderMania was inspired by the repetitive process of finding suitable tender opportunities. Many companies still need to manually check multiple procurement portals, read tender requirements one by one, and filter out tenders that are not relevant to their capabilities. This wastes time and often makes teams miss better opportunities.

What it does

TenderMania is an autonomous AI agent that searches for tender opportunities, normalizes the data, scores each tender, and generates a draft Expression of Interest. It evaluates tenders based on company capability, eligibility, and win probability. Only tenders that pass the scoring and rule-based filters are recommended for human approval.

How we built it

We built TenderMania using Python, OpenAI-compatible LLMs, Pydantic, BM25 RAG, and SQLite. The system uses multiple agents: scraper agents collect tender data, a normalizer agent structures the information, scorer agents evaluate each tender, an aggregator decides whether to pursue or archive it, and a drafter agent creates a personalized EOI draft. The final step is protected by a human approval gate.

Challenges we ran into

One of the biggest challenges was making sure the agent did not recommend irrelevant tenders. Tender data can be messy, incomplete, or unrelated to the company’s actual capabilities. Another challenge was making the generated EOI draft specific and useful instead of generic. We also had to balance LLM reasoning with deterministic rules so the system stayed reliable and safe.

Accomplishments that we're proud of

We are proud that TenderMania can run an end-to-end autonomous workflow from tender discovery to draft generation. We also built multi-dimensional scoring, hard-gate filtering, RAG-grounded drafting, and an audit trail that records the reasoning behind each decision. This makes the system more transparent and trustworthy.

What we learned

We learned that agentic AI works best when LLM reasoning is combined with structured data, validation, and clear rules. We also learned how important explainability is in business workflows. For tender decisions, it is not enough for an AI to say “this is a good fit”; it must explain why, using evidence from the company profile and tender requirements.

What's next for TenderMania

Next, we want to add scheduled daily tender scanning, integrate more procurement portals such as LPSE, and support approval through Slack or WhatsApp. We also plan to build a dashboard for companies to manage tender opportunities, track win/loss outcomes, and continuously improve the scoring system over time.