Projects
47 active projects across teams · 12 archived · 8 cross-org collaborations
Quarterly recertification ahead of May 21 expiration. Faithfulness regression remediation v17→v18, comprehensive red-team coverage, FREE-AI evidence, EU AI Act Art. 9-15 documentation refresh.
Migrating from Bedrock Agents to multi-cloud (Bedrock + Vertex) with framework-neutral policy layer.
MCP tool surface hardening — image-based prompt injection coverage and tool-misuse defense.
DPDP Act algorithmic-software due-diligence cycle preparation for AML investigation agent.
Expanding customer-support-rag with Hindi & Hinglish coverage; dataset curation and judge-model selection.
First Indic-native voice agent certification submission with Sarvam-1 as primary.
Baseline evaluation for branch operations knowledge agent.
Quarterly regression sweep on internal helpdesk copilot.
SR 11-7 quantitative validation of mortgage disclosure generation with independent challenger model.
Caste/religion/region bias regression sweep; remediation A/B in progress.
EU AI Act Art. 11 technical documentation refresh for Frankfurt-based wealth explainer.
Migration from in-house orchestration to Vertex Agent Builder for pension guidance.
Migration to Bedrock AgentCore with comprehensive red-team re-baselining.
Multi-agent CrewAI architecture governance baseline; cross-agent escalation policies.
Quarterly DORA-aligned vendor reassessment for card dispute classifier vendor.
Umbrella initiative coordinating Q3 FREE-AI evidence collection across all in-scope artifacts.
Frankfurt evidence pack assembly for EU AI Act Article 9-15 across high-risk artifacts.
Annual DPDP cycle — DPIAs, automated-decision disclosure, retention reviews.
Preparation for ISO 42001 Q4 surveillance audit; gap analysis underway.
Cycle reassessment of all critical AI vendors per DORA.
Frontier model assessment for potential primary upgrade across claims and wealth.
Q3 frontier model sweep — claude-3-7, gpt-5, gemini-2.5, llama-4.
Indic-native model evaluation across Hindi/Hinglish/Tamil/Bengali coverage.
Side-by-side: LangGraph, CrewAI, Bedrock Agents, Vertex Agent Builder, Copilot Studio.
Cross-org collaboration: localize claims-copilot for EU and US insurance arms.
Joint investment-research-assistant program between EU asset mgmt and SG HQ.
Joint frontier model evaluation between Banking India and Innovation Lab.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.
Internal tooling baseline.