--- title: Data Science & AI Portfolio – Credit Risk, RAG, and Multi-Agent Systems emoji: 📊 colorFrom: blue colorTo: indigo sdk: gradio sdk_version: "4.44.0" app_file: app.py pinned: true license: mit short_description: Data Science and AI portfolio --- ## Professional Summary Data Scientist with deep expertise in credit risk modeling, statistical analysis, and advanced AI—including agentic systems and Retrieval-Augmented Generation (RAG) architectures. Demonstrated success developing scalable, interpretable models for multi-billion-dollar portfolios and leading analytics innovations for capital adequacy and scenario-driven loss forecasting. Proficient in Python, SQL, and state-of-the-art machine learning, with a strong track record in conducting rigorous quantitative research, enhancing predictive reliability, and delivering actionable insights that support high-stakes financial decision-making. --- ## Education **Rutgers Business School – Master of Quantitative Finance** _New Jersey, USA • 2019 – 2021_ - Relevant Coursework: Fixed Income, Derivatives, Econometrics, Financial Time Series, Stochastic Calculus, Optimization Models, OOP (C++/Python), Risk Management, Equity Trading **BITS Pilani – B.Tech in Electrical & Electronics Engineering, Minor in Finance** _Hyderabad, India • 2013 – 2017_ --- ## Experience ### Global Monitor – Data Scientist _Hyderabad, India • October 2024 – Present_ - Developed and optimized conversational AI systems enabling dynamic engagement with market research data, driving a **30% increase in actionable insight extraction**. - Engineered NLP pipelines and integrated advanced question-answering models to transform complex business datasets into intuitive responses, **reducing analysis turnaround and accelerating data exploration speed by 40%**. - Collaborated with analysts to design AI-powered tools leveraging databases of **10M+ records** for market trends, sustainability insights, and benchmarking, while applying **Agentic AI and RAG techniques** to autonomously enhance retrieval, reasoning, and precision of responses from large-scale datasets. --- ### CRISIL (an S&P Global Company) – Data Scientist _New York, USA • March 2021 – July 2024_ - Designed a loan pricing framework by segmenting portfolios into behavioral clusters using borrower-level features; applied unsupervised learning and regression analysis to tailor pricing strategies, enhancing **risk-adjusted returns** and **model interpretability**. - Built a comprehensive **credit risk model for a \$5Bn cryptocurrency loan portfolio**, incorporating PD, LGD, EAD, EEPE, Pluto–Tasche methodology (non-margin lending), and credit VaR calculations. - Enabled **scenario-driven loss forecasting** by developing models with tunable key risk variables; improved transparency, adaptability, and predictive reliability of capital adequacy assessments, contributing to a **six-month delay in bankruptcy**. - Engineered a **dual Monte Carlo simulation** framework projecting 365-day price paths and a stress-testing variant with 1.5× negative return weighting to increase robustness under adverse market conditions. - Migrated and re-engineered four financial models from R, MATLAB, and SQL into Python using object-oriented design and scientific libraries; achieved measurable improvements in **accuracy**, **speed**, and **maintainability** for multi-million-dollar analytics products. - Automated data pipelines and reporting tools with Python for **25+ Tier-1 banks**, interpreting technical documentation to create modular scripts that increased throughput and enabled quantitative teams to focus on higher-value activities. --- ### Deutsche Bank – Credit Risk Analyst _Mumbai, India • February 2017 – June 2019_ - Analyzed business model risks through financial statement analysis, liquidity profiling, and peer benchmarking across a portfolio of **40 companies**, enabling dynamic rating adjustments and early risk detection. - Built and implemented credit scorecard models for corporates, banks, insurers, and brokerages in Southeast Asia by integrating financial ratios, macroeconomic indicators, and sector-specific risk factors. - Developed forward-looking financial projections combining firm KPIs, strategic plans, macroeconomic trends, and discounted cash flow (DCF) valuation inputs. - Delivered accurate monthly credit assessments for financial institutions and conducted comprehensive peer analysis and financial research using Bloomberg, broker reports, and public filings to support rating decisions and continuous credit monitoring. --- ## Focused Skill Advancement (2024 – Present) Following my tenure at CRISIL, I took a deliberate personal break to recalibrate and focus on **high-impact emerging technologies**. Recognizing the transformative potential of Artificial Intelligence, I invested in rigorous industry certification programs and hands-on development work, with emphasis on: - Large Language Models (LLMs) and **agentic systems** - **Retrieval-Augmented Generation (RAG)** architectures - Applied AI in **quantitative finance** and risk analytics This period has been dedicated to building **real-world projects** and production-ready systems, ensuring my expertise bridges deep theoretical knowledge with practical implementation. --- ## Projects ### NyayaSpashti – AI-Powered Legal Research Assistant _Hugging Face Space • Generative AI, RAG, Legal NLP_ Link: [`NyayaSpashti on Hugging Face`](https://huggingface.co/spaces/the-dreimar/NyayaSpashti) - Architected a **Generative AI legal assistant** for 12 core Indian Acts using a **hybrid RAG pipeline** combining Chroma semantic search, BM25 keyword matching, and cross-encoder reranking to maximize retrieval accuracy. - Engineered a secure and robust **Google Gemini** integration with advanced retry logic, rate-limit management, and guardrails to prevent off-topic responses and unauthorized legal advice. - Designed and implemented a **self-verifying RAG workflow**: intelligent document chunking, hybrid retrieval, LLM generation, automated verification, and section-level source citations. - Deployed a **production-ready application on Hugging Face Spaces**, utilizing Git LFS for index versioning and secure API key management. - Built a dynamic **PDF-to-citation mapping** system ensuring every response includes exact source references and table names, significantly improving explainability and user trust. --- ### Credit Default Prediction (American Express) _Kaggle • Credit Risk, Gradient Boosting, Explainability_ Link: [`Credit Default Prediction Notebook`](https://www.kaggle.com/code/thuggilisaireddy/credit-default-prediction) - Developed a **LightGBM model** to predict credit card default probabilities on an **11M+ record dataset**, executing extensive EDA, feature engineering, and hyperparameter tuning. - Applied robust validation techniques including **TimeSeriesSplit** and **StratifiedKFold**. - Evaluated model performance with **AUC** and **Gini**, and used **SHAP** and **LIME** for model interpretability and stakeholder communication. --- ### MarketLens – Multi-Agent Financial Analysis _Hugging Face Space • Multi-Agent Systems, Finance, LLMs_ Link: [`MarketLens on Hugging Face`](https://huggingface.co/spaces/the-dreimar/MarketLens) - Architected a **multi-agent investment system** using CrewAI with six specialized agents (Fundamental Analyst, Technical Analyst, News Sentiment Analyst, Bull/Bear Researchers, Trader) for comprehensive stock evaluation and automated trade recommendations. - Designed **iterative debate workflows** between opposing research agents to surface multi-perspective analysis and reduce confirmation bias. - Built a production data pipeline integrating **SEC EDGAR API** with XBRL parsing to extract financial statements for S&P 500 companies and compute **15+ key ratios** for automated fundamental analysis. - Integrated **Google Gemini** with a map–reduce strategy for SEC 10‑K/10‑Q summarization, including robust retry logic and rate-limit handling. - Developed an NLP pipeline to extract critical sections from SEC filings with **95%+ accuracy**, and aggregated real-time market data from multiple sources, reducing API overhead by **60%** via intelligent caching. - Deployed a full-stack **Streamlit application on Hugging Face Spaces**, delivering AI-generated trade action reports with buy/hold/sell recommendations and detailed investment analysis. --- ## Certifications - **AWS Certified AI Practitioner (AIF‑C01)** – Foundational expertise in AI/ML and generative AI deployment on AWS (2025). _Credential: AWS AI Practitioner (Credly)_ - **IBM Certificate in Retrieval-Augmented Generation (RAG) and Agentic AI** – Advanced RAG pipelines and multi-agent AI frameworks (2025). _Credential: IBM RAG & Agentic AI (Coursera)_ - **Applied Data Science Lab – WorldQuant University** – Intensive applied analytics program on real-world financial datasets (2024). _Credential: WorldQuant University Applied Data Science (Credly)_ --- ## Technical Skills - **Programming & Databases**: Python, SQL (PostgreSQL, MySQL), R, MATLAB - **AI/ML Frameworks**: LangChain, CrewAI, Hugging Face, Google Gemini, OpenAI, RAG Architectures, Multi-Agent Systems - **Machine Learning & Model Development**: Supervised Learning (Classification, Regression), Unsupervised Learning (Clustering, PCA), Ensemble Methods (Random Forest, Gradient Boosting), Deep Learning, Time Series Forecasting, Cross-Validation, Hyperparameter Tuning, Model Interpretability (SHAP, LIME), Feature Engineering - **Financial Modeling**: Credit Risk Modeling (PD, LGD, EAD, VaR), Monte Carlo Simulation, Stress Testing, Scenario Analysis - **Cloud & MLOps**: AWS (AI Practitioner Certified), MLflow, GitHub Actions, CI/CD, Docker - **Visualization & Deployment**: Streamlit, Plotly, Matplotlib, Seaborn, Hugging Face Spaces --- ## Contact - **Hugging Face**: [`the-dreimar`](https://huggingface.co/the-dreimar) - **Selected Spaces**: - [`NyayaSpashti`](https://huggingface.co/spaces/the-dreimar/NyayaSpashti) – AI Legal Research Assistant - [`MarketLens`](https://huggingface.co/spaces/the-dreimar/MarketLens) – Multi-Agent Financial Analysis If you’re interested in advanced AI for credit risk, financial analytics, and RAG/agentic systems, I’d be happy to connect.