Building data systems
for quantitative finance

Architecting high-performance data systems for front-office quants and traders across Spot, Options, and Swaps desks.

• Launched AI initiative using Claude SDK + MCP, enabling natural language queries across internal data
• Scaled real-time PnL & risk systems processing ~3M trades/day via vectorized Pandas + multiprocessing
• Built ETL pipelines with Airflow & Jenkins transforming raw data into post-trade analytics
• Designed FastAPI risk transfer system with dry-run impact analysis and approval workflows
• Optimized MongoDB & ClickHouse with query guardrails to prevent DoS from expensive queries

Built the internal MLOps platform powering model development, deployment, and monitoring across the bank.

• Automated model retraining/redeployment, reducing manual intervention by 90%
• Built backend services + core DB schemas with Python & Django for unified ML platform
• Implemented GitOps with ArgoCD for Spark jobs on Kubernetes, supporting department-wide model deployments

Built trading systems and NLP pipelines for the proprietary trading desk, supporting credit sentiment analysis and platform modernization.

• Built NLP pipeline for credit sentiment: NER + disambiguation + FinBERT for pricing distressed bonds
• Led Front Arena upgrade with performance improvements and vendor coordination (Certified Scrum Master)
• Led migration from on-prem Solaris to Linux VPC, improving stability and maintainability