Feb 2024 – Present · 2 yrs 4 mos · Navi Mumbai · Hybrid

• NL2SQL for Equity Rating DB: Built a maintainable Natural Language → SQL system with 83%
• evaluation accuracy; presented to C-suite stakeholders.
• Increased robustness by schema+domain context injection and systematic refinement for quality improvement.
• (Athena, Azure OpenAI, Pinecone, LangChain, RAG techniques: column-based search, keyword-optimized vector
• retrieval, few-shot prompting, revision feedback loop)
• Attribute Resolution Scalability (7→14 million): Designed and executed production-like scalability tests;
• identified memory ceilings from in-pod caching and provided concrete capacity and autoscaling
• recommendations. (Kubernetes, Flux, Locust, Prometheus/Grafana, EFK)
• Attribute Resolution Re-architecture: Rearchitected service to externalize cache from kubernetes pods to
• an external database to enable reliable performance at 14 million scale; accomplished 85% cost reduction, 97%
• faster startup, 64% more throughput, less errors, all while maintaining model quality. (Redis in ElastiCache
• Valkey, Postgres in RDS)
• Risk Model Smart Text LLM Integration: Productionized infrastructure and deployment workflows for
• integrating LLM output into Morningstar Direct. (AWS, Harness CI/CD, Terraform, Docker)
• Investor Pulse Model Analysis: Delivered accuracy metrics and explainable drivers to stakeholders for
• decision-making on continuation of the product, saving resources on maintenance. (Python, statsmodels)
• Marketing Analytics Data Foundation: Implemented Google Analytics 3 data backup and performed EDA
• to support acquisition/churn modeling. (AWS S3, Python, SQL)
• CI/CD Performance and Platform Maintenance: Reduced CI runtime by 66%; maintained and secured
• multiple production rating services; migrated Disaster Recovery pipeline and standardized artifacts. (Harness,
• SonarQube, Terraform, Jenkins→Harness, msnexus→Sonar artifacts)