About JanAvlokan
An AI-powered decision-support platform enabling transparent, accountable, and efficient welfare delivery across government schemes.
Our Mission
Government welfare programs in India serve millions daily across food security, education, energy, and employment schemes. Due to the massive scale, diversity of schemes, and regional variations, traditional audit mechanisms struggle to monitor misuse in real time.
Public finance studies and Comptroller and Auditor General (CAG) reports estimate that 20-40% of subsidy value is lost to inefficiencies or leakage across large-scale schemes.
JanAvlokan addresses this critical need for a scalable, privacy-preserving, and explainable intelligence system that can flag high-risk patterns early, assist administrators in prioritizing audits, and preserve fairness for genuine beneficiaries.
Vision Statement
To transform reactive audits into proactive governance intelligence, strengthening transparency and public trust in India's welfare delivery system.
Core Principles
Built on foundations of privacy, transparency, and ethical AI
Privacy First
No personally identifiable information enters the cloud. All sensitive identifiers are irreversibly hashed.
Transparency
Every flagged case comes with human-readable explanations for administrative review and audit defensibility.
Fairness
Advisory-only system that never blocks or delays welfare payments to genuine beneficiaries.
Human-in-the-Loop
Final decisions always rest with human administrators. AI provides intelligence, not verdicts.
Active Learning
Auditor feedback (Verify/Dismiss) is recorded and used to retrain ML models, continuously improving detection accuracy.
Welfare Schemes Monitored
Current deployment covers major national welfare programs
PM POSHAN
Mid-Day Meal Scheme
Pradhan Mantri Poshan Shakti Nirman (revamped 2021)
Children in Classes I–VIII in government & aided schools
- Inflated beneficiary counts
- Duplicate student enrollments
- Ghost schools and fictitious claims
- Unusual meal distribution patterns
PMUY + PAHAL
LPG Subsidy Schemes
Pradhan Mantri Ujjwala Yojana (2016) & Direct Benefit Transfer for LPG (2014)
BPL/SECC households receiving LPG connections and subsidies
- Duplicate LPG connections
- Fake beneficiary accounts
- Excessive refill frequency
- Shared bank account patterns
System Capacity
Implementation Phases
System deployment and operationalization
Data Integration
Connected to PFMS and DBT transaction streams
Model Training
Trained unsupervised models on anonymized historical data
System Deployment
Live deployment with real-time monitoring enabled
Full Operations
Operational across multiple schemes and regions
Support and Assistance
Need help using JanAvlokan? Our support team is here to assist you.
Submit a Support Request
support@janavlokan.gov.in
For technical assistance
Ministry of Electronics and IT
New Delhi, India
+91-11-XXXX-XXXX
Mon-Fri, 9:30 AM - 5:30 PM IST
Frequently Asked Questions
Is JanAvlokan currently deployed?
JanAvlokan is currently a prototype developed as part of the Hack-4Viksit Bharat initiative. We are seeking partnerships for pilot deployments with state governments.
How does JanAvlokan protect beneficiary privacy?
All personally identifiable information is irreversibly hashed before processing. No PII enters our cloud infrastructure and all outputs are advisory-only with human decision-making.
Can JanAvlokan integrate with existing systems?
Yes, JanAvlokan is designed to work as an advisory layer over existing DBT/PFMS systems without requiring changes to payment infrastructure.
What schemes can JanAvlokan analyze?
The platform is scheme-agnostic and can be configured to analyze any welfare program that generates transactional data, including Pradhan Mantri Ujjwala Yojana, PM POSHAN, and more.
What types of fraud patterns does JanAvlokan detect?
JanAvlokan identifies five key fraud risk categories: Unusual Activity, Suspicious Locations, Scheme Overlaps, Beneficiary Clusters, and Repeat Withdrawals. Each category represents a different mode of potential fraud or system misuse.
Can I upload my own data for scanning?
Yes. The CSV Quick Scan feature lets you upload beneficiary transaction data in CSV format. The system validates the file, runs ML inference via the Vertex AI endpoint, and instantly returns per-row risk levels and flags.
How are audit reports generated?
JanAvlokan includes a full Report Builder with a collaborative rich-text editor. You can link flagged transactions as evidence, add findings with severity ratings, and export finalized reports to PDF or DOCX for official submission.
Does the system notify officials automatically?
Yes. Automated Email Alerts are sent to district-level officials via the Gmail API when high-risk anomalies are detected. Alerts include a risk summary, flagged beneficiary details, and direct links to the dashboard.
Does auditor feedback improve the AI?
Absolutely. The Audit Panel includes a feedback loop—officers can verify fraud or dismiss false positives. This feedback is stored and used to retrain models on Vertex AI, enabling Active Learning and continuously improving detection accuracy.
Ready to Get Started?
Access the dashboard to view real-time risk assessments, explore flagged cases, and generate audit reports for your schemes and jurisdiction.