At Holiday Channel Inc, I deployed a production RAG service with intent-based query routing, citation-grounded responses, and confidence-gated abstention. I also improved Recall@5 by 22% with a hybrid retrieval pipeline for a 50K+ SKU catalog.
I engineered ETL ingestion pipelines using Auto Loader and distributed PySpark UDFs, maintaining data freshness with sub-1s latency.
At Avaya, I built and owned Node.js and Spring Boot microservices for a CCaaS platform supporting Kafka workloads in a distributed system processing 1M+ messages per day. I also improved service reliability and expanded CI pipelines with tests that raised coverage to 80% on critical modules.
On my CUDA-Accelerated VLM Fine-Tuning & Inference Engine project, I assembled and distributed a multimodal model across T4 GPUs. I authored fused C++ CUDA/Triton kernels and used INT8 quantization to achieve a 3x inference speedup.
