Problem
1: AI driven Credit for borrower with no credit history
Many individual
persons and MSMEs in India do not have credit history to
obtain loans from banks in the traditional way. They are
deprived of loans despite genuine need, while banks are unable
to provide loans due to lack of credit history and risk of
NPAs.
Expected Outcome: Develop an
AI‑powered alternate credit scoring system that leverages
diverse data sources such as phone bill payment consistency,
e‑commerce purchase behavior, geolocation stability,
questionnaire‑based risk assessment, merchant ratings, and
bank account cash flow patterns (if available). The prototype
should include: (1) Psychometric & behavioral risk models, (2)
Consent‑based data flow mechanisms, and (3) Privacy &
encryption compliance. This solution should enable underserved
communities to access loans digitally without traditional
credit history, while ensuring responsible lending practices.
Problem
2: Audio Forensics for Voice Security
These days
Generative AI tools can now clone a customer’s voice by
capturing just few seconds of audio. Fraudsters use this to
bypassVoice Biometric passwords or call center agents
into transferring funds. Current defenses rely on metadata
(phone numbers), which is easily spoofed.These days Generative
AI tools can now clone a customer’s voice by capturing just
few seconds of audio. Fraudsters use this to bypass Voice
Biometric passwords or call center agents into transferring
funds. Current defenses rely on metadata (phone numbers),
which is easily spoofed
Expected Outcome: A real-time
Audio Forensics Module that analyzes the spectrogram (visual
representation of audio) of a live call. It must detect
synthetic artifacts—micro-imperfections in pitch and
frequency that human ears miss but machines can spot—and flag
the call as High Risk within the first 10 seconds.
Problem
3: Automated Discovery of Misconfigurations in Open Source
Dependencies
Modern banking
applications rely on hundreds of opensource libraries, often
left unpatched or misconfigured, creating hidden security
risks. Existing SCA tools mainly detect known CVEs but fail to
identify insecure defaults like hardcoded passwords or weak
cipher suites. Developers miss vulnerabilities during fast
releases, expanding the attack surface. The challenge is to
build a static analysis tool that scans code and dependencies
to detect both CVEs and insecure configurations. It must
integrate into CI/CD pipelines, provide clear reports, and
suggest automated fixes to enforce secure-by-design practices.
Expected Outcome: The
solution should automatically detect vulnerable dependencies
and insecure default configurations across code and
configuration files, while providing clear, actionable fix
suggestions or even auto-remediation options for developers.
It must integrate seamlessly into the CI/CD pipeline, ensuring
secure-by-design development practices without impacting
delivery speed.