Aishwarya Pawar
List of Publications (Papers, Book Chapters, Reviews, etc.)
Time Series Analysis: Bitcoin and S&P 500 Data
Data Analysis Using Statistical Methods
Work Experience(s)
Data Scientist
• Developed deep learning pipelines (PyTorch) for predictive maintenance and anomaly detection on sensor data; leveraged NNs and LSTMs to achieve 93.8% accuracy
• Deployed ML services via Docker and Flask on AWS EC2/Lambda, delivering real-time predictive insights to clients; introduced MLflow for reproducibility and monitoring, cutting debugging time by 25%
• Applied NLP with BERT embeddings and PySpark pipelines (cutting processing time by 35%) to classify and cluster financial transactions, enabling customer segmentation and fraud-risk detection
Quant Finance Research Intern
• Researched equity strategies using company fundamentals and income statement data, applying factor models to enhance stock selection across cross-sector investments and portfolio optimization
• Developed sector, price, and volume-based trading screens, implemented back-tests and evaluated risk-adjusted returns
• Leveraged Value at Risk (VaR), Monte Carlo simulations, and ML models (Random Forest, LSTM) to predict market trends, assess risks, and evaluate portfolio returns, supporting data-driven investment strategies
ML Engineer (Software Engineer)
• Collaborated to build a survey intelligence pipeline using LLM embeddings and RAG retrieval over government/NGO datasets, fine-tuning BERT classifiers for demographic tagging
• Delivered low-latency inference APIs with FastAPI, Docker to host predictive models, improving response time by 40%
• Created MLOps lifecycle with MLflow, CI/CD, and SageMaker, deploying production models with 99.9% uptime
Data Analyst (Credit and Risk)
• Designed, backtested, and deployed credit score binning models with SHAP/LIME explainability; containerized via Docker and deployed on AWS, boosting funding volume by 40%
• Applied AI/ML, statical and NLP (BERT) models to link borrower behavior with macroeconomic indicators; built risk-optimized funding grids using LGD and PD models that reduced portfolio exposure by 6% across $250M in assets
• Forecasted leads using time-series models (Prophet, SARIMA) and Media Mix Model to optimize spend allocation
• Engineered automated reporting pipelines with SQL and Power BI dashboards, delivering real-time KPI tracking for 120+ metrics used by C-suite execs for credit and risk decisions
AI Engineer (Data Scientist)
• Architected an LLM-powered chatbot that ingested large enterprise documents, stored embeddings in FAISS, and applied summarization (Flan-T5, Llama-2 fine-tuned); deployed via FastAPI on AWS, reducing support resolution time by 35%
• Built predictive models (XGBoost, Random Forest) for churn and growth decomposition; guided a targeted retention campaign whose A/B test showed a 9% reduction in churn
• Developed customer segmentation models with clustering (k-means) and enhanced via embedding-based retrieval (RAG) and LoRA-tuned BERT models, improving customer classification precision and enabling targeted outreach
Teaching Experience(s)
Graduate Teaching Assistant
• Assisted students with Applied Time Series Analysis and Probability Theory concepts, created and graded HWs and exams