Jyoti SHARMA
Faridabad
Summary
Data Scientist with 4 years of hands-on industry experience delivering scalable machine learning, deep learning, and Generative AI solutions that drive real business value.
Overview
5
5
years of professional experience
1
1
Certification
Work History
Data Scientist
OrBEX Technologies PVT LTD
07.2021 - Current
- Designed and developed an enterprise-scale intelligent knowledge assistant using a hybrid RAG architecture to deliver accurate, context-aware question answering over large organizational document repositories.
- Architected a complete RAG pipeline covering document ingestion, intelligent chunking, embedding generation, hybrid retrieval (dense vector + sparse keyword/BM25), re-ranking, and LLM-driven response generation.
- Implemented hybrid search techniques to balance semantic relevance with exact-match accuracy for domain-specific terms, identifiers, and policy references.
- Built metadata-aware retrieval mechanisms with role-based and contextual filters (department, document type, time range) to ensure secure and relevant responses.
- Integrated LLMs using structured prompts, context windows, and citation logic to generate grounded, explainable answers with source attribution.
- Developed scalable REST APIs using FastAPI with asynchronous processing to support low-latency query handling.
- Delivered a reliable and scalable enterprise knowledge platform that significantly improved information accessibility and reduced reliance on manual document searches.
- Improved answer relevance and factual accuracy through the combination of semantic and keyword-based retrieval techniques.
- Developed an enterprise-scale analytics solution to predict disease progression risk for chronic patients using machine learning, combined with a retrieval-augmented generation (rag) system to provide clinical explanations and guideline-based recommendations.
- Collected and analyzed patient-level clinical data including demographics, diagnoses, lab results, vitals, and historical medical records.
- Conducted exploratory data analysis (EDA) to identify trends, disease patterns, and key predictors of disease progression.
- Built and evaluated machine learning models (Random Forest, XGBoost) to predict disease progression risk scores.
- Optimized model performance using hyperparameter tuning and evaluated results using metrics such as ROC-AUC, precision, recall, and F1-score.
- Implemented a RAG-based clinical knowledge support system to retrieve relevant treatment guidelines, clinical protocols, and research summaries from internal medical documents.
- Integrated ML predictions with RAG outputs to provide contextual explanations such as contributing risk factors and recommended monitoring actions.
- Developed APIs using FastAPI to expose prediction and clinical explanation services.
- Collaborated with healthcare stakeholders to translate clinical requirements into data-driven solutions.
- Created dashboards and reports to visualize risk distribution, model performance, and patient risk trends.
- Enabled early identification of high-risk patients using machine learning–based disease progression risk scoring.
- Improved clinical decision-making by providing guideline-backed explanations through a RAG-based clinical knowledge system.
- Developed an end-to-end Machine Learning–based Credit Risk Scoring system to assess the probability of loan default and support data-driven lending decisions.
- Worked closely with business stakeholders to understand lending workflows, credit approval criteria, and regulatory constraints.
- Defined the target variable as loan default (binary classification) and framed the problem as a probability-based risk scoring task rather than a hard accept/reject decision.
- Identified key business objectives such as minimizing defaults, controlling false approvals, and maintaining regulatory compliance.
- Collected and analyzed historical loan application data including customer demographics, income, employment details, credit history, and repayment behavior.
- Performed extensive Exploratory Data Analysis (EDA) to identify missing values, outliers, skewed distributions, and multicollinearity among features.
- Analyzed default rate trends across customer segments to identify high-risk patterns.
- Built baseline models using Logistic Regression to ensure interpretability and establish benchmark performance.
- Converted predicted probabilities into interpretable credit risk scores and risk bands (Low / Medium / High).
- Improved loan default detection by accurately identifying high-risk applicants before approval.
- Reduced financial risk by minimizing risky loans.
Education
Bachelor of Technology(B.Tech) -
Siddhi Vinayak College Of Engineering
Alwar, Rajasthan, IndiaSkills
- Supervised and Unsupervised Machine Learning techniques
- Statistical Analysis
- Modeling
- Hypothesis Testing
- Feature Engineering
- Data Preparation for ML models
- Predictive Modeling
- Advanced Analytics
- Experiment Design
- A/B Testing
- Data Visualization
- Reporting
- Interactive Dashboards
- Natural Language Processing
- Text Classification
- Sentiment Analysis
- End-to-End Model Deployment
- APIs
- Flask
- FastAPI
- Streamlit
- SQL Query Optimization
- Data Analysis
- Big Data Processing
- Business Problem Framing
- Requirement Analysis
- Cloud-based Machine Learning Solutions
- AWS
- Azure
- Convolutional Neural Networks
- ResNet
- EfficientNet
- U-Net
- YOLO
- R-CNN
- Sequence Models
- RNN
- LSTM
- GRU
- Seq2Seq Architectures
- Transformer Models
- BERT
- GPT
- ViT
- T5
- Attention Mechanisms
- Sequence Learning
- Generative Models
- GANs
- Diffusion Techniques
- Model Optimization Techniques
- Quantization
- Pruning
- Knowledge Distillation
- Large Language Models
- LLaMA
- Mistral
- Prompt Engineering
- Structured Prompt Design
- Retrieval-Augmented Generation
- RAG
- Vector Databases
- Pinecone
- FAISS
- Chroma
- Programming
- Data Handling
- Python
- SQL
- Pandas
- NumPy
- Matplotlib
- Seaborn
- Scikit-learn
- TensorFlow
- Docker
- LLM-based systems
- Hugging Face APIs
- LangChain
- Parameter-efficient fine-tuning
- Retrieval-Augmented Generation Pipelines
- Backend Frameworks
- Creative thinker
- Ability to communicate
- Ability to manage multiple projects
- Team worker
- Collaborative
- Responsible
Certification
- IBM Certified Machine Learning
- Google Cloud Professional Data Engineer
Personal Information
- Date of Birth: 07/20/88
- Nationality: Indian
- Marital Status: Married
Languages
- Hindi
- English
- Hindi
Timeline
Data Scientist
OrBEX Technologies PVT LTD
07.2021 - Current
Bachelor of Technology(B.Tech) -
Siddhi Vinayak College Of Engineering