H
HemanandaSaiSimhachalamNaidu Singampalli
About
Detail
Boca Raton, Florida, United States
Résumé
Jobs
verified_user
0% verified
GenAI EngineerMovn HealthJan 2025 - Current (1 year 10 months)• Designed and built agentic AI systems leveraging multi-agent workflows, tool-using LLMs, and stateful memory architectures to enable autonomous reasoning, planning, and decision-making in production environments.
• Developed intelligent agents using LangChain, LangGraph, and OpenAI's Agents SDK, implementing function-calling LLMs, dynamic tool routing, dependency-aware agent orchestration, and integrations with external APIs.
• Architected Retrieval-Augmented Generation (RAG) pipelines using semantic search, embedding models, and vector databases (Pinecone, FAISS), tightly coupled with agent execution flows for grounded, context-aware responses.
• Improved response accuracy and trustworthiness by implementing source attribu- LAI/ML EngineerLentra AI PVT LtdFeb 2023 - Dec 2023 (11 months)• Led AI-driven credit decision systems, combining traditional ML models with LLM-assisted feature analysis and decision explainability, improving decision accuracy by 15%.
• Built scalable data and feature pipelines from transactional and behavioral data, serving as inputs for ML models and downstream AI-driven decision workflows, reducing data preparation time by 40%.
• Optimized and orchestrated ensemble ML models (Random Forest, Gradient Boosting, XGBoost) using cross-validation and Bayesian tuning, improving AUC from 0.75 to 0.85 and supporting AI-driven decision pipelines.
• Deployed AI inference services on AWS using Docker and Kubernetes (EKS), supporting real-time decision scoring with latency under 100 ms.
• Bu - ML EngineerCognizant Technology SolutionsJan 2020 - Feb 2023 (3 years 2 months)• Developed and deployed production ML models for critical business use cases such as risk scoring, churn prediction, and fraud detection using Python, scikit-learn, and XGBoost.
• Built and trained deep learning solutions using TensorFlow/Keras for image classification, OCR, and sequence modeling within insurance and finance.
• Designed and fine-tuned NLP pipelines for text classification, entity recognition, and sentiment analysis using spaCy, NLTK.
• Implemented semantic search and QA systems using TF-IDF, BM25, and early dense retrieval models such as SBERT and DPR.
• Delivered ML services using FastAPI/Flask, containerized with Docker and deployed via Azure ML and CI/CD pipelines.
• Applied model explainability
Education
verified_user
0% verified
- Master in Data ScienceFlorida Atlantic UniversityJan 2024 - Dec 2025 (2 years)
- SSQL: Data Reporting and Analysis
- FFull Stack Web Development
- EExcel Black Belt L4
- PPower BI Data Analyst Associate
- PPython for Machine Learning & Data Science
- AAZURE Certified Cloud Practitioner
Projects (professional or personal)
verified_user
0% verified
- DDistributed and Optimized LLM Fine-Tuning Pipeline• Architected a distributed training pipeline to fine-tune Mistral 7B on a domain-specific dataset using DeepSpeed and FSDP for efficient model and optimizer sharding.
• Reduced training time by optimizing parallelism strategies and resource utilization, enabling scalable fine-tuning of large models.
• Integrated MLflow to evaluate scaling efficiency across configurations and deployed the fine-tuned model via a high-throughput inference server leveraging continuous batching. - SSmart Academic Form & Process Assistant (AI Agent)• Engineered a multi-step conversational AI agent to automate academic form submission workflows using tool invocation and function calling.
• Designed agent orchestration to convert natural-language user inputs into structured JSON payloads and system-generated emails, replacing manual, email-based processes.
• Implemented cross-session state tracking and proactive reminders, significantly reducing form errors and improving submission accuracy for time-sensitive tasks. - EEnd-to-End ML Pipeline (User Churn Prediction)• Architected and deployed a production-ready ML pipeline on Azure for user churn prediction, covering data ingestion, feature engineering, training, and real-time inference.
• Built a high-performance classification model using LightGBM, optimizing model accuracy and inference latency.
• Integrated MLflow for experiment tracking, parameter logging, and artifact management to ensure reproducibility and controlled model iteration.