CRPF Tender Evaluation

Nyayadarsi

AI-Powered Tender Evaluation & Accountability Engine

Restoring the integrity of the L1 mandate and providing ironclad legal protection for honest evaluation officers.

Explore Pipeline Meet the Team
1. Fulfilling the Vision

Defending the Process

Evaluating officers face immense pressure. We look beyond standard document parsing to solve the two fundamental threats to procurement integrity.

Accidental Non-Compliance

A legitimate small business submits a valid certificate under a slightly different legal name. A rigid automated OCR system eliminates them.

This is a failure of the tool.

Deliberate Obfuscation

A sophisticated external bad actor buries a fraudulent financial certificate in a degraded scan, actively trying to trick a human evaluator under deadline pressure.

This is an attack on the process.
2. The Core Solution

Restoring Integrity to L1

Financial bids remain strictly masked until eligibility is verified. We sequence decision-making to structurally prevent manipulation.

Phase 1

Document Eligibility (AI)

Incoming bids are processed through our ingestion engine. Bidders who clearly fail mandatory criteria are auto-excluded with a full citation trail.

Phase 2

The YELLOW Queue (Human + AI)

The AI does not make silent legal decisions. It triages ambiguous cases into a structured queue, presenting the exact extracted anomaly and empowering the officer to log a digitally signed pass/fail decision.

Phase 3

Micro-Collusion Check

Before final award, the system runs cross-document analysis on the submitted cohort to detect shared metadata or forensic similarities indicating cartel behavior.

3. Technical Architecture

Built for Reality

Standard OCR destroys table structures. Nyayadarsi is built specifically for the reality of complex Indian procurement documents.

Format Detection

Identifies digital PDFs, scanned images, and structured tables.

Layout-Aware Extraction

Utilizes LayoutLMv3 to preserve the spatial relationships of financial tables and technical matrices.

The Financial Ontology

Uses a curated, domain-specific mapping of Indian CA certificates to identify valid equivalents without relying on exact, rigid keyword matches.

Evaluation Engine Confidence Scoring

Verdicts are reached using an optimized LLM running purely in extractive mode to prevent hallucination.

Score Band Confidence System Action
GREEN 85–100% Evidence verified. Auto-marks eligible.
YELLOW Below 85% Ambiguous. Triggers mandatory human review with a structured reason statement.
RED 85–100% (fail) Value clearly below threshold. Auto-marks ineligible with full citation chain.
4. The Audit Trail

Legal Protection for Officers

Honest officers who make correct disqualification decisions frequently face legal petitions. Every automated action and human override generates an append-only log:

  • Microsecond Precision: Timestamp to microsecond precision.
  • Cryptographic Hash: Hash of the source document to prove it wasn't tampered with.
  • Verbatim Extraction: Values cited by page and cell location.
  • Accountability: Officer Digital Signature ID for any human override.
5. Mitigation

Risks & Limitations

LLM Hallucination

Mitigated by restricting the LLM strictly to extractive QA. It is forbidden from inferring numbers.

Sophisticated Forgery

Mitigated through cross-document consistency checks (e.g., turnover on a CA certificate must logically match the submitted ITR).

Legal Ambiguity

Subjective criteria like "works of similar complexity" are flagged as YELLOW rather than guessed, keeping legal authority firmly with the human officer.

6. Beyond Evaluation

The Roadmap

Phase 2 (Pre-Tender)

An "Integrity Alert" system that analyzes drafting criteria to prevent restrictive specifications.

Phase 3 (Post-Award)

A builder dashboard requiring daily, GPS-verified progress uploads analyzed by computer vision to authorize milestone payments.

Execution-First Environment

Team Coding_Aghoris

This is an execution-first battleground. The organizers don't want theoretical whitepapers; they want working prototypes solving specific, real-world constraints. This is where socially driven architecture thrives.

Suryansh

Sibam

Pritam

Satya

Timeline

April 10 — May 4, 2026

Venue

Taj Yeshwantpur, Bengaluru

Ultimate Prize

Pitch at PAN IIT Summit (May 16) to VCs & Policymakers

"Nyayadarsi does not replace the procurement officer. It empowers them. It gives the honest officer the documentation to do their job efficiently and the undeniable evidence to defend every decision. This is India's procurement accountability infrastructure."