Agentic AI & RAG Systems Specialist

Mragisha Jain

Raleigh, NC +1-919-***-**** *******@****.*** linkedin.com/in/mragisha-jain github.com/mragisha EDUCATION

North Carolina State University,Raleigh, NC, USA Aug 2024 - May 2026 Master of Science, Computer Science (Data Science specialisation) GPA- 3.89/4 Coursework - Deep Learning, Algorithms, Neural Networks, Foundation of Data Science, Database Management System Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal, India Aug 2018 - July 2022 Bachelor of Technology, Computer Science and Engineering GPA- 3.75/4.0 SKILLS

Languages: Python, Java, SQL, C++, Ruby, JavaScript, R. AI/ML: LangChain, LangGraph, LlamaIndex, PyTorch, TensorFlow, RAG, Agentic AI, RLHF, VectorDB, NLP, HuggingFace,Ollama. Backend & Big Data: Spring Boot, Apache Kafka, Apache Spark, Hadoop, RESTful APIs, Microservices, Ruby on Rails. Cloud & DevOps: AWS (EC2, S3), Jenkins, Grafana, Prometheus, Docker, Kubernetes, CI/CD, Git. Tools: ReactJS, R Shiny, Tableau, PowerBI, Salesforce Cloud, MariaDB, PostgreSQL. WORK EXPERIENCE

Envestnet Yodlee Raleigh, NC

Data Science Intern June 2025 - Aug 2025

● Architected an Agentic AI orchestration system using LangGraph and MCP (Model Context Protocol) with human-in-the-loop workflows, leveraging DSPy to programmatically optimize prompt signatures, improving tool execution efficiency by 35%.

● Implemented self-correcting LLM pipelines using Qwen3 via Ollama, vLLM; utilized Databricks for scalable data processing and managed the model lifecycle to monitor faithfulness, reducing query failure rates by 42%.

● Designed an LLM-powered insights service with a ReactJS frontend and asynchronous backend processing, enabling scalable delivery of personalized financial recommendations.

AIAL Lab - North Carolina State University Raleigh, NC Research Assistant Aug 2025 - Present

● Built a production-grade RAG system using OpenAI embeddings, FAISS vector search, and transformer-based cross-encoder reranking, achieving 90%+ intent classification accuracy with less than 3s end-to-end latency.

● Designed a multi-stage hybrid retrieval pipeline combining dense and sparse search, MMR-based diversity filtering, and source boosting, improving retrieval precision by 40% compared to baseline keyword-only systems.

● Implemented LLM-powered intent classification and entity extraction with GPT-4 using few-shot prompting, replacing rule-based logic with semantic understanding and significantly improving topic identification and downstream response quality. Camcore NCSU Raleigh, NC

Machine Learning Research Assistant Jan 2025 - Present

● Co-authored TreeSuit, a climate-based species recommendation tool (currently under review), utilizing bioclimatic variables

(BIO1–BIO19) and CMIP6 projections to simulate species selection across global climate horizons.

● Developed a multi-index suitability model integrating GBIF, TerraClimate, and CHIRPS datasets; optimized environmental classification through clustering (K-Means, DBSCAN) to automate breeding decision scenarios. Persistent Systems Limited Pune, India

Software Engineer July 2022 – July 2024

● Architected a microservices-based loan platform using Java, Spring Boot, and Apache Kafka, automating credit assessment and scaling monthly disbursements from $6M to $35M.

● Streamlined delivery cycles by orchestrating Jenkins and Docker-based CI/CD pipelines, achieving 90% unit test coverage and a 40% reduction in deployment time.

● Implemented an observability suite using Grafana and Prometheus to monitor API latency, proactively resolving 95% of production bottlenecks during peak transaction cycles.

● Designed secure REST APIs for identity verification, leveraging Spring Security and AWS S3 to ensure encrypted document storage and 99.9% system uptime.

PROJECTS

Comparing Network Pruning and Knowledge Distillation Deep Learning

●Benchmarked Network Pruning vs. Knowledge Distillation on VGG and ResNet architectures under FLOP constraints; analyzed sparsity patterns to optimize latency and memory for edge deployment. Speech Emotion Recognition LSTM, GRU

●Developed BiGRU models using MFCC features for speech emotion recognition, achieving 90% accuracy; optimized training pipelines in TensorFlow/Keras with dropout and Adam tuning.

Medical - Customer Support Chatbot RLHF, DPO

● Developed a RAG-based medical chatbot using LangChain, OpenAI GPT-3.5, and FAISS, implementing semantic search over 47K+ doctor–patient conversations with conversation memory to deliver context-aware, multi-turn clinical responses.

● Built an end-to-end RLHF pipeline using Direct Preference Optimization (DPO) with Hugging Face TRL, fine-tuning Llama 3.2 using QLoRA (4-bit quantization + LoRA adapters) to efficiently align model behavior on domain-specific preference data, reducing GPU memory usage while maintaining response quality.

Contact this candidate