N
Nitish Kumar Manthri
About
Detail
United States
Résumé
Jobs
verified_user
0% verified
AI/ML EngineerAmerican ExpressAug 2024 - Current (1 year 10 months)• Architected RAG-based AI Risk Copilot (LangChain, GPT-4, LLaMA) integrating transaction history and merchant risk signals; enabled 90K+ fraud/credit analyst reviews with full audit traceability • Deployed Pinecone vector database indexing 7M+ records with BERT embeddings; achieved sub-100ms retrieval latency and reduced LLM hallucinations by 35% through semantic grounding • Engineered multi-agent LLM workflows with chain-of-thought prompting for case summarization, anomaly reasoning, and automated policy citation; cut analyst review time by 40% • Designed LLM evaluation framework (RAGAS metrics, hallucination detection, grounding checks) with human-in-the-loop validation; achieved 95% response accuracy pre-deployment • Implemented Respons- Machine Learning EngineerCVS HealthSep 2023 - Jul 2024 (11 months)• Programmed transformer-based clinical NLP (BERT/BioBERT) extracting structured indicators from 500K+ physician notes; achieved 90%+ F1-score for condition tagging • Established predictive risk models (XGBoost, LightGBM) on 5M+ claims/EHR records; reduced high-risk cohort misclassification by 19% and prioritized 50K+ care interventions • Configured A/B testing framework for pharmacy engagement (2M+ interactions); drove 15% improvement in medication adherence saving estimated $2M annually • Formulated Azure Data Factory/Airflow ETL pipelines with automated validation; reduced data refresh latency by 30% while maintaining HIPAA compliance • Rolled out models via MLflow with containerized APIs on Azure ML; implemented automated drift detectio
- KData ScientistKPIT TechnologiesJun 2020 - Aug 2022 (2 years 3 months)• Developed credit risk models (XGBoost, Random Forest) on 200K+ records achieving 78% precision; reduced false positive alerts by 15% saving $250K in manual review costs • Orchestrated LSTM sequence models (TensorFlow) analyzing 12-18-month repayment histories; improved early default detection by 8% identifying $1.2M in at-risk accounts • Formed behavioral features (rolling repayment ratios, utilization trends, delinquency indicators); improved high-risk account recall by 11% • Created PyTorch NLP pipelines for customer service classification (0.83 F1); automated routing of 30K+ monthly dispute cases • Released batch scoring APIs (FastAPI/Docker) with Airflow-orchestrated retraining and MLflow tracking; reduced model refresh cycle from 2 w
Education
verified_user
0% verified
- M.S. Computer ScienceUniversity of MemphisMay 2024Coursework: Deep Learning, NLP, Neural Networks, Big Data Analytics
- AWS Certified ML SpecialtyAmazon Web Services
- TensorFlow Developer CertificateGoogle
- Databricks Data Engineer AssociateDatabricks
- Google Cloud Professional ML EngineerGoogle Cloud
- NVIDIA Certified Professional (GenAI)NVIDIA
- Azure Data Scientist AssociateMicrosoft
- Databricks ML AssociateDatabricks
- Azure AI Engineer AssociateMicrosoft
- NVIDIA Certified Associate (GenAI & LLMs)NVIDIA