Back to all projects Tender Intelligence · Indian Railways

Tenderintel — tenders, bids & pricing, all in one mind.

Tender-management software purpose-built for Indian Railways procurement. Tenderintel automates the tedious — logging in to the portal, downloading comparative statements, extracting bidder pricing — and turns it into clean analytics our clients can act on.

ClientNeetu Yoshi Pvt Ltd
DomainIndian Railways procurement
StatusActive development
StackPython · FastAPI · React · Playwright · SQLite · Gemini AI

The problem.

Indian Railways runs one of the largest public-procurement systems in the world through IREPS — thousands of tenders, hundreds of items, dozens of bidders per lot. For a supplier like Neetu Yoshi, staying competitive means understanding who's bidding, at what prices, and for which items — data that lives scattered across PDFs behind a CAPTCHA-protected portal.

The old workflow was all manual: download each comparative statement by hand, transcribe bidder pricing into spreadsheets, and pray nobody mistyped. Hours lost every week, and by the time you had the data, the next tender was already closing.

What we built

A quiet, always-on tender brain.

Tenderintel is a full pipeline: automated scraping of the IREPS government portal, an AI-powered PDF parser for comparative statements, a tidy relational store of price history, and a modern React dashboard for analysis.

  • Authenticated scraping: Playwright handles login, DSC prompts, CAPTCHA (via 2Captcha), and OTP flows, maintaining session cookies across long runs.
  • Comparative-PDF parsing: a combined pdfplumber + Gemini pipeline turns messy multi-page statements into structured bidder records — rate, charges, GST, rank, and more.
  • Item catalog & mapping: every tender is tied back to our item catalog (Centre Pivot Bottom, Wedge, Brake Beam, Axle Box Housing, Bogie Bolster, Side Frame, Spring Plank and more) via explicit mapping rules.
  • Price history & vendor types: a normalised price_history table lets analysts see how each bidder's rates trend across zones, consignees, and time.
  • Streamlit + React workspace: operators trigger data-collection runs from a simple UI; analysts explore the data in a polished React dashboard built on shadcn/ui and TanStack Table.
  • Email orchestration: structured templates and a sender module handle the communications side — the same intelligence layer that runs the data also runs the outreach.
In numbers

Built to handle real procurement volume.

500MB+Base catalog size
DozensActive tender items tracked
Hours → MinutesManual parsing time saved
100%Audit-traceable records
What's next

From analytics to decisions.

The next chapter is decision intelligence. We're layering in bid-price recommendations, win-rate forecasts, and anomaly alerts when a new entrant disrupts an item's pricing. The goal isn't just to see the market clearly — it's to know exactly what to bid, and why.

Talk to us about tender intelligence