At Mphasis (Client: Citi), I refactor backend PySpark and Hive SQL transformation jobs to address performance bottlenecks, data skew, and calculation anomalies across enterprise datasets.
I’ve also engineered automated processing and validation pipelines for data governance, deletion workflows, and lifecycle retention compliance. I created synthetic data frameworks and SQL reconciliation tests to check schema drift and transaction edge cases.
At Ielektron Technologies (Client: Magna Global), I engineered an AWS-based perception validation pipeline comparing DUT models with multimodal Ground Truth using Hungarian bipartite matching. I also helped reduce edge-case false negatives by about 15% across retraining cycles through error taxonomy and hard-example mining splits.
In my projects, I’ve built multi-agent systems for AML alert investigation and job screening, and a local tabular profiler using DuckDB and a fine-tuned LLaMA 3.1 model. I deployed quantized GGUF models via Ollama to generate visualization artifacts from statistical profiles.
