Open to opportunities

kaushal shivaprakash

@kaushalshivaprakash

Message

I’m a data engineer turning large-scale event and ML pipelines into trusted, production-ready data.

United States

Message

What I'm looking for

I’m looking to own production data and MLOps pipelines—especially data quality gates, near real-time observability, and automation that keeps ML inputs clean and auditable. I want teams that value engineering rigor and measurable reliability.

I’m a Data Engineer focused on end-to-end ownership of data collection, quality, and MLOps pipelines on cloud platforms. I build systems that keep datasets clean, validated, and ready for model training—without letting bad inputs silently reach production.

At Centific (Client: Amazon), I designed and owned AWS data collection and quality pipelines for Ring camera and home device programs, serving as the sole data quality gate across 500+ collection stations. I created validation and metadata observability layers on AWS S3 and Athena to detect defects, missing capture signals, and station-level health issues in near real time, cutting investigation time from days to hours.

I’ve also built event-driven ingestion and analytics infrastructure at Cognizant (Client: JPMorgan Chase & Co), including Apache Kafka + PySpark pipelines and production Airflow orchestration for ML feature stores and analytics workloads. I redesigned a Redshift warehouse into an optimized star schema and implemented automated dbt and Python data quality checks to prevent schema regressions over 2+ years.

As a Graduate Research Assistant at University at Buffalo (UBCDS), I delivered Azure Machine Learning training pipelines and MLflow-tracked experiments, improving preprocessing time and model performance for carbon footprint estimation. I bring a research-driven mindset to production engineering—automating quality, lineage, and operational reliability.

Experience

Work history, roles, and key accomplishments

Current

Data Engineer

Current

Centific

Mar 2026 - Present (5 months)

Designed and owned an end-to-end AWS data collection and quality pipeline for Amazon Ring camera/home device programs, acting as the sole data quality gate before SageMaker fine-tuning. Built validation and near-real-time observability on AWS to prevent defective or incomplete data from entering model training and replaced manual hardware readiness checks across 500+ stations.

Python SQL MLOps AWS Sagemaker Data Pipelines Predicate Pushdown Data Observability Fine Tuning

Graduate Research Assistant

University at Buffalo

Aug 2024 - Jan 2026 (1 year 5 months)

Built end-to-end ML training pipelines on Azure Machine Learning for UB Campus Dining and Shops to model carbon footprint emissions from 1M+ meal recipe and ingredient records. Trained and tracked XGBoost, Random Forest, and K-Means models in MLflow, improving prediction/clustering accuracy versus a logistic regression baseline.

Python Azure Machine Learning PySpark MLFlow Airflow XGBoost Random Forest K Means Broadcast Joins Model Registry

Software Engineer 2

Cognizant Technology Solutions

Jan 2022 - Aug 2024 (2 years 7 months)

Built Kafka and PySpark event-driven ingestion pipelines at JPMorgan Chase scale, supporting 50K+ events/sec and maintaining 99.95% end-to-end consistency. Redesigned Redshift into a star schema to reduce query latency, orchestrated 50+ Airflow DAGs for production reliability, and implemented automated dbt-based data quality checks feeding ML-ready, audit-ready datasets to stakeholders.

Python Kafka PySpark Airflow Amazon Redshift AWS DBT Docker Power BI Star Schemas Data Quality Testing Data Lineage

AgentEval AI

Microsoft Agents League

Jan 2026 - Present (7 months)

Built an agentic AI evaluation platform for Microsoft Copilot using LangGraph and Azure OpenAI with a RAG pipeline backed by pgvector. Implemented evaluator agents to detect reasoning drift and perform automated root-cause analysis to surface quality signals before deployment.