AI/ML & Generative AI Engineer

KRUPAL PATEL

******.*@************.*** +1-551-***-**** NJ, United States LinkedIn

SUMMARY

AI/ML Engineer with 4+ years of experience turning raw data and research ideas into working systems that people

actually use. The path started with computer vision and NLP projects for healthcare and insurance clients, where the

focus was on making messy documents and images produce clear, usable answers. That early work grew into deeper

experience with generative AI, including retrieval-augmented search, multi-agent workflows, and large language model

fine-tuning, applied in a regulated manufacturing and life sciences environment. Comfortable moving across the full

lifecycle, from training and evaluating models to packaging them into APIs and monitoring their behavior once they

reach production. Strong grounding in cloud platforms, vector databases, and MLOps practices, paired with an eye for

practical outcomes like faster response times, fewer manual reviews, and measurable cost savings.

SKILLS

Programming Languages: Python, SQL, Java, C++

AI/ML Frameworks: TensorFlow, PyTorch, Scikit-learn, Keras, XGBoost, ONNX

Generative AI & LLMs: GPT-4/GPT-4o, Claude, Gemini, LangChain, LlamaIndex, RAG Architecture, Prompt Engineering,

LLM Fine-Tuning, Hugging Face, Vector Embeddings, Semantic Search

AI Agents & Orchestration: LangGraph, AutoGen, CrewAI, Multi-Agent Systems, Function Calling, Tool Integration

Vector Databases: Pinecone, ChromaDB, Weaviate, FAISS, Milvus

MLOps & CI/CD: MLflow, Docker, Kubernetes, GitHub Actions, Jenkins, Terraform, Model Registry, Model Monitoring

Cloud Platforms: AWS (Bedrock, SageMaker, Lambda, EC2, S3), Azure OpenAI Service, Google Cloud (Vertex AI)

Data Engineering: Apache Spark, Kafka, Airflow, ETL/ELT Pipelines, Databricks, Snowflake

NLP & Computer Vision: spaCy, NLTK, OpenCV, YOLO, Named Entity Recognition, Image Segmentation

Backend & APIs: FastAPI, Flask, REST APIs, GraphQL, Microservices Architecture

Databases: PostgreSQL, MongoDB, MySQL, Redis

Data Analysis & Visualization: Pandas, NumPy, Tableau, Power BI, Matplotlib

Testing & Optimization: A/B Testing, Model Evaluation Metrics, Hyperparameter Tuning, Quantization, Latency Optimization

Collaboration Tools: Git, GitHub, Jira, Confluence, Agile/Scrum

EXPERIENCE

Thermo Fisher Scientific Inc, NJ, USA AI/ML Engineer Jan 2026 - Present

• Implemented a semantic search layer over internal lab notebooks using ChromaDB and vector embeddings, helping scientists

locate prior experiment results 3x faster and reducing duplicate testing efforts.

• Streamlined a FastAPI-based microservice for real-time inference on quality control data, handling roughly 50,000 requests per

day while keeping average response time under 200 milliseconds.

• Applied quantization techniques to reduce a production computer vision model's size by 60%, enabling deployment on edge

devices in the lab without sacrificing detection accuracy.

• Coordinated A/B testing frameworks to compare two candidate demand forecasting models, identifying a version that improved

forecast accuracy by 14% before full-scale rollout.

• Introduced a function-calling based tool integration layer connecting an internal chatbot to SAP and inventory systems, cutting

supply chain query resolution time by roughly 30%.

Mphasis, India AI/ML Engineer Jan 2021 - Aug 2024

• Constructed a named entity recognition model using spaCy and NLTK to extract key terms from insurance claim documents,

shortening manual claims review time by 33% for a US-based insurance client.

• Managed data storage and retrieval workflows across PostgreSQL and MongoDB for a healthcare analytics project, supporting

a dataset that grew to over 8 million patient records without performance degradation.

• Set up a Kafka-based streaming pipeline to feed real-time fraud detection alerts, decreasing average alert latency from 5 minutes

to under 45 seconds for a payment’s client.

• Refined an image segmentation model with TensorFlow and Keras for a healthcare imaging use case, improving diagnostic

flagging accuracy by 21% during internal validation testing.

• Wrote infrastructure automation scripts using Terraform and Jenkins to provision cloud environments for model training,

cutting environment setup time from three days to under four hours.

EDUCATION

Master of Science in Computer Science

Montclair State University, NJ, United States

Bachelor of Engineering in Computer Engineering

Sai Institute of Technology & Engineering Research, India

PROJECTS

AI PDF Chatbot (Mini RAG) – Python, Streamlit, LangChain, FAISS, OpenAI

• Built a Retrieval-Augmented Generation (RAG) chatbot that enables users to upload PDF documents and receive accurate,

context-aware answers using semantic search and LLM-powered responses.

• Implemented document parsing, text chunking, vector embeddings, and FAISS similarity search to retrieve relevant context

before generating responses with source citations.

• Developed an interactive Streamlit interface, reducing document search time by over 80% through AI-powered natural

language querying.

AI Travel Planner (Multi-Agent System) - React, FastAPI, LangGraph, CrewAI, OpenAI, Google Maps API, Weather

API, Docker

• Developed a multi-agent AI application that generates personalized travel itineraries by coordinating specialized agents for

flights, hotels, weather, budgeting, and attractions.

• Integrated LLM reasoning, tool calling, and external APIs using LangGraph/CrewAI to automate end-to-end travel planning

with dynamic recommendations.

• Built a conversational full-stack application capable of itinerary optimization, budget estimation, and real-time travel

assistance through intelligent agent collaboration.

Enterprise Multi-Document RAG Assistant - React, FastAPI, LangChain, LangGraph, OpenAI, Pinecone, Docker

• Built a production-grade Retrieval-Augmented Generation (RAG) platform enabling users to query PDFs, Word documents,

and text files using natural language with citation-backed responses.

• Implemented document ingestion, semantic chunking, vector embeddings, Pinecone vector search, metadata filtering, and

conversational memory to improve retrieval accuracy.

• Developed a scalable React + FastAPI application with streaming responses, authentication, and Docker deployment,

reducing enterprise document search time through AI-powered semantic retrieval.

Contact this candidate