At Swiftforce.AI, I designed and shipped a HIPAA-compliant LLM service using GPT-4 and LangChain, serving 1,000+ daily interactions at sub-200ms p95 latency. I also built an event-driven ETL pipeline with evaluation and observability hooks, reaching 95% intent-classification accuracy.
I own MLOps across AWS, where Redis caching cut inference costs 30%. CI/CD and MLflow/OpenTelemetry monitoring supported 99.9% uptime, with PHI encryption, RBAC, and JWT authentication in place.
I conceived, built, and launched Vastra, a consumer AI wardrobe app. It combines Gemini outfit-photo analysis with a custom JavaScript astronomy engine that computes planetary positions within 0.01° of reference charts.
My project work also includes fine-tuning HRNet-W48 for pose estimation and building a multimodal model that fused LiDAR and visual data for 6G wireless networks. At Sensegood Instruments Pvt. Ltd., I built a genetic-algorithm optimizer and an automated ETL pipeline.
