Saurav Kapadiya
+91 9016981176 kapadiyasaurav2003@gmail.com Mehsana, Gujarat
github kaggle linkedin
Objective
AI/ML Engineer with proven experience in building scalable GenAI applications using LLMs (BERT, OpenAI, etc.), LangChain,
and vector databases. Skilled in developing RAG pipelines, prompt engineering, and deploying ML services via CI/CD on AWS.
Eager to solve real-world problems using cutting-edge NLP and GenAI solutions.
Experience
Thinkbiz Technology Pvt Ltd May 2024 – Present
AI/ML Developer Ahmedabad, Gujarat, India
• Developed LLM-based invoice automation system using OCR + OpenAI, reducing manual effort by 70%.
• Built and deployed RAG pipelines using Qdrant and OpenAI embeddings for document search.
• Built CI/CD pipelines using Docker + GitHub Actions, and deployed services using AWS Lambda.
Institute for Plasma Research Jan 2024 – May 2024
Deep Learning Intern Gandhinagar, Gujarat, India
• Built LSTM-based model to forecast plasma disruptions with 92% accuracy, contributing to early warnings in nuclear
fusion research.
• Optimized model performance using Keras & TensorFlow and handled end-to-end ML pipeline for research insights.
Education
Dhirubhai Ambani Institute of Information and Communication Technology (DA-IICT) Nov 2020 - May 2024
B.Tech in Information and Communication Technology- ICT Gandhinagar, Gujarat
CGPA: 6.2/10
Projects
Copilot – RAG-based Chatbot for Company Policies: LangChain | OpenAI | RAG | Qdrant | Semantic Search| AWS
• Developed an AI-powered Q&A chatbot that answers natural language queries based on uploaded company policy
PDFs using RAG (Retrieval-Augmented Generation) architecture.
• Used LangChain with OpenAI GPT and Qdrant vector database for chunked embedding storage and semantic search.
• Prompt Enginnering and fallback mechanisms to ensure reliable responses even in ambiguous queries.
• Deployed the backend using FastAPI, with full integration into the company’s production-grade product
infrastructure.
Expense Automation: LangChain, OpenAI, PostgreSQL, AWS Lambda, Prompt Engineering
• Implemented GenAI-based invoice processing workflows using Langchain and LLMs.
• Automated Japanese enterprise compliance validation using rule-based and ML techniques.
• Designed and automated test cases for ML pipelines using PyTest, enabling early-stage defect detection.
Accounting Automation: ML, BERT, Qdrant Vector Database, Langchain, AWS Lambda
• Built predictive models using BERT for financial classification tasks; achieved 98%+ accuracy.
• Applied pattern discovery techniques and semantic similarity (vector-based clustering).
Plasma Current Disruption – Time Series Forecasting: LSTM, BiLSTM, GRU, Transformer, Attention Mechanism, TensorFlow
• Applied deep learning-based time series forecasting in a scientific research environment with critical impact.s
• Developed and compared multiple deep learning models (LSTM, BiLSTM, GRU, Transformer) to forecast plasma
current disruptions using time-series data from fusion reactors.
• Managed the full ML lifecycle from data preprocessing and model building to evaluation and reporting.
• Integrated attention mechanisms and encoder-decoder architectures to enhance model interpretability and long-
range dependency capture.
• Achieved significant performance improvement over classical forecasting baselines; contributed to early warning
systems for experimental plasma control.
AI Mock Interviewer: OpenAI | Whisper | Google TTS | LangChain
• Designed a GenAI-based mock interview platform where users upload their CV and job description, and the system
generates personalized interview questions using LLMs.
• Integrated Whisper (speech-to-text) to capture user responses and Google TTS for voice-based question delivery,
creating a realistic voice-interaction experience.
• Enabled feedback and analytics for users to assess performance and prepare for real-world interviews.
Commute Expense Automation System: LangChain Agents | OCR | Navitime API | PostgreSQL | OpenAI
• Built a system to automate commute reimbursement workflows for employees using public or private
transportation, in line with company policy.
• Used OCR for extracting data from required documents (licenses, tickets, passes) and LangChain multi-agent
architecture to handle tasks like distance calculation, route optimization, and policy validation.
• Stored validated data in PostgreSQL, ensuring traceable and compliant decision-making.
Technical Skills
Programming Languages: Python, C/C++
ML Frameworks: TensorFlow, Keras, PyTorch, Scikit-learn, OpenCV, Pandas, Numpy, Matplotlib
LLMOps & Vector Search: Qdrant, FAISS, Weaviate, Embedding-based retrieval, OpenAI Embeddings, BERT base Tokenizer
Test Automation & CI/CD: PyTest, GitHub Actions, Docker, Unit Testing
Databases & Querying: PostgreSQL, MySQL, DynamoDB, NoSQL, SQLite
Cloud & DevOps: AWS Lambda, S3, CloudWatch, ECR, CodePipeline, CodeBuild, Docker, Step Functions, Parameter Store,
Secret Manager, RDS, KMS
AI/ML: Machine Learning, Deep Learning, Computer Vision, RNN, CNN, LSTM, Encoder-Decoder models, Transformer
models, BERT, AI/ML Integration, OpenAI APIs, Time Series Forecasting, Gen AI, Prompt Engineering
Version Control: Git, GitHub, GitLab, GitHub Actions, Bitbucket
LLMs & GenAI: OpenAI Apis, Prompt Engineering, LangChain, LLM Agents, Fine-tuning of BERT Models, RAG pipelines, BERT
Fine-tuning
Certifications
Supervised Machine Learning: Regression and Classification: DeepLearning.AI
