DEEKSHA GANDHI PATTEM

Developed end-to-end Generative AI and RAG pipelines on AWS Bedrock using LangChain and LangGraph to provide context-aware tax guidance for 2M+ users and reduce average resolution time by 43%. Built and deployed multi-step AI agents using prompt engineering and tool-use orchestration to automate financial document classification, cutting manual review time by 57% across 5 product lines. Integrated Python/FastAPI microservices on cloud-native architecture (Kubernetes, Docker) with Apache Kafka for event streaming, supporting 60K+ concurrent requests at sub-180ms p95 latency with 99.8% uptime. Contributed to MLOps workflows and model deployment pipelines via CI/CD, reducing model release cycles from 2 weeks to under 2 days. Implemented vector database indexing with Hugging Face embeddings across 12M+ financial documents, boosting semantic retrieval precision by 32%. Applied scikit-learn and ML models for fraud detection achieving 95.1% precision and preventing $3.4M in fraud losses.

Developed and maintained 10+ RESTful APIs powering 3 core SaaS modules for 18K+ daily active users, reducing average page load time by 36%. Designed a microservices-based backend with Python/Flask and Docker containers orchestrated on Kubernetes, cutting service provisioning time by 68% and enabling zero-downtime deployments. Built end-to-end CI/CD pipelines using GitHub Actions and Docker, accelerating release cadence from bi-weekly to daily deployments and reducing post-release defect rate by 54%. Optimized normalized PostgreSQL schemas and tuned queries for high-throughput transactional workloads, improving analytical query performance by 51% across 10 production tables exceeding 25M rows. Integrated third-party REST APIs (payment gateways, SMS, OAuth2) and built a unified API gateway, reducing cross-service latency by 29% and external API failure rate by 41%. Built responsive React.js dashboards with reusable component libraries, reducing UI development time by 38% and improving mo