Infrastructure Automation

• Built a C#-based observability system with exception classification and real-time email/messenger notifications, for faster incident detection and triage acceleration. Introduced mute rules for noisy exceptions to reduce alert fatigue and enabled one-click bug creation from exception stack traces in the UI via Azure DevOps integration

• Engineered a scalable data pipeline to ingest and analyze 600K+ tweets and 10K+ user profiles using LLM-based sentiment classification, with automated data retrieval and storage to support investment research; enabled repeatable execution and export of results via CSV and REST API for integration with downstream research tools (Excel, dashboards, etc.)

• Containerized a multi-service portfolio dashboard (R/Shiny UI + backend, Python REST API) using multi-stage Docker builds, reducing image size from 1.8 GB to 620 MB; configured Docker Compose with persistent volumes and isolated networks for secure inter-service communication

• Introduced an Azure DevOps CI/CD pipeline to build container images on a dedicated agent and deploy them to a separate production server; the pipeline was later adopted as a template for 3 additional projects to standardize and streamline deployments across the team

• Architected and deployed an invoice automation system that ingests invoices from email, extracts structured data using LangChain-powered LLMs, performs deterministic and LLM-based reconciliation against internal records, and auto-posts transactions to the accounting system, eliminating ~2.5 hours of manual work per week

• Delivered Kafka-based structured logging for AI Gateway, an OpenShift-hosted LLM router used firmwide; added Prometheus metrics for component latency tracking, and enabled incremental LLM responses via Server-Sent Events (Python/Flask) to avoid blocking users

• Developed a C# chatbot backend using Semantic Kernel with RAG on Redis-stored embeddings of 100+ internal documents; built for wealth managers, it auto-detects when to reference documentation or invoke one of 15 APIs based on user intent

• Integrated OpenAI APIs to develop an automated summarization service that analyzed transcripts from 200+ meetings, generated concise recaps, and distributed them via email to cross-functional teams to enhance information sharing and reduce manual note-taking overhead

• Mentored an intern through structured code reviews, pair programming sessions, and Agile retrospectives, leading to a 60% improvement in her story completion rate (from 8 to 13 stories per sprint) and significantly accelerating her development as a productive team contributor

• Developed a Pytest test suite (integration and unit tests) for a Python project, resulting in 21% decline in bugs

• Consolidated 15 microservices into Istio service mesh architecture to simplify service-to-service communication

• Abstracted out authorization policy enforcement for 15 microservices using Open Policy Agent’s Policy-as-code

• Achieved seamless authorization policy enforcement by implementing a Sidecar pattern through Envoy Proxy