N
Nithiya Varsha Markandan Rajendren
About
Detail
Virginia, United States
Résumé
Jobs
verified_user
0% verified
- BData ScientistBlue Horse DigitalSep 2025 - Current (11 months)• Designed and evaluated a document processing and retrieval system using Azure OpenAl and Azure Al Search, defining evaluation metrics and optimizing chunking and retrieval strategies to improve contextual relevance while reducing latency by 70%. • Developed and optimized semantic chunking and retrieval strategies, reducing LLM token usage by 35% (~$2.4K/month) while preserving contextual relevance across large document sets. • Built and validated a hybrid retrieval system (vector embeddings and semantic ranking), achieving 91% precision through iterative tuning of chunking and retrieval parameters. • Conducted performance analysis of end-to-end pipelines, implementing reliability mechanisms (async processing, retries, monitoring) to maint
Al EngineerAllwyn CorporationJan 2025 - May 2025 (5 months)• Developed an enterprise RAG-based document intelligence system using AWS Bedrock, Kendra, Lambda, and S3, improving information retrieval precision to 92% across 1,500+ documents and enabling automated proposal workflows. • Designed and evaluated embeddings-based retrieval pipelines, optimizing document representation and ranking strategies to improve relevance and reduce document miss rates. • Built a prompt evaluation and benchmarking framework (Python, Streamlit) to compare multiple LLMs (Claude, Llama, Jamba) across 31 enterprise queries, improving response relevance by 50% through iterative experimentation. • Conducted model and prompt performance analysis, implementing validation checks and prompt versioning strategies to improve ou- Technical AnalystBank of AmericaSep 2021 - Jun 2023 (1 year 10 months)• Developed and deployed real-time fraud detection models (XGBoost, Logistic Regression) on high-volume transaction data (5M+ events/day), improving recall by 21% while maintaining precision through threshold optimization. • Engineered predictive features, including transaction velocity, merchant risk scoring, and time-dependent behavioral patterns, and evaluated feature impact using precision-recall metrics and offline validation datasets. • Designed a hybrid decision system combining machine learning predictions with rule-based logic, enabling real-time risk scoring and contributing to ~$3.2M annual fraud loss prevention. • Conducted iterative model evaluation by comparing model variants and threshold strategies, analyzing tradeoffs betwe
Education
verified_user
0% verified
- GMaster of Science in Data Analytics EngineeringGeorge Mason University, VirginiaAug 2023 - May 2025 (1 year 10 months)GPA 3.9/4.0
- Bachelor of Engineering in Computer Science and EngineeringSRM Institute of Science and TechnologyJun 2017 - May 2021 (4 years)GPA 3.6/4.0
Projects (professional or personal)
verified_user
0% verified
- AAWS Powered COVID-19 Data Lake Analytics PipelineBuilt a serverless data pipeline using AWS to automate CSV ingestion and transformation, enabling fully managed ETL workflows. Queried data via Athena and created QuickSight dashboards, reducing manual reporting by 80%. Orchestrated real-time processing and monitoring with CloudWatch for a scalable, production-ready pipeline.
- YYouTube Trending Analytics PipelineCreated an end-to-end cloud pipeline to analyze trending videos via the YouTube Data API (Python on SageMaker), stored in Parquet, landed in S3, and modelled in BigQuery with a STAR schema. Published interactive Tableau dashboards and designed a roadmap for refresh automation and Al-generated daily summaries.