Nawaraj Paudel, PhD

Led design and development of an AI-assisted fraud investigation platform for financial crime analysts. The platform triages alerts, surfaces evidence across transactions, devices, and auth events, and uses an agentic AI pipeline to recommend dispositions — Account Takeover, Synthetic Identity, Identity Theft, or Non-Fraud. Built the full stack: React/TypeScript frontend with PwC Appkit 4, FastAPI/PostgreSQL/Redis backend, and LangGraph agent pipeline with LiteLLM for multi-provider LLM inference. 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝗲𝗱 𝗮𝗻 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 using LangGraph and LiteLLM that generates evidence summaries and outcome recommendations. Designed with configurable privacy modes, multi-provider inference (Azure OpenAI, OpenAI, Ollama), and automatic failover for production-safe AI in a regulated financial environment. 𝗕𝘂𝗶𝗹𝘁 𝗮 𝘁𝗵𝗿𝗲𝗲-𝘁𝗶𝗲𝗿 𝗰𝗮𝗰𝗵𝗶𝗻𝗴 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 (self-replenishing hot pool, Redis, PostgreSQL) delivering sub-second load times for high-priority alert bundles. A background daemon precomputes top-N highest-risk bundles with agent results, transactions, and rule hits. 𝗗𝗲𝘀𝗶𝗴𝗻𝗲𝗱 𝗮 𝘀𝘆𝗻𝘁𝗵𝗲𝘁𝗶𝗰 𝗳𝗿𝗮𝘂𝗱 𝗱𝗮𝘁𝗮 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 with scenario injectors (ATO, Synthetic Identity, Identity Theft) generating realistic customer profiles, accounts, transactions, auth events, and device telemetry. Built a deterministic rule engine to produce alerts, enabling end-to-end testing without production data. 𝗗𝗲𝗳𝗶𝗻𝗲𝗱 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁𝘀 with global Transaction Monitoring and KYC teams, translating AML/fraud domain logic into engineering specs and a policy-driven rule engine. Helped establish cloud infrastructure connections to accelerate parallel workstreams. 𝗠𝗲𝗻𝘁𝗼𝗿𝗲𝗱 𝗷𝘂𝗻𝗶𝗼𝗿 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀 on agentic AI (LangGraph state graphs, tool-use patterns), ML fundamentals, network analysis, and full-stack best practices across Python and TypeScript.

Developed end-to-end ML solutions including production-ready models, interactive apps (using Streamlit, FastAPI), and LLM-powered chatbots. Built AI dental assistant for godental.ai using RAG architecture, FAISS, and GenAI. Delivered custom solutions across healthcare, real estate, and finance sectors using Python, SQL, R, and modern MLOps practices (Git, Docker, MLFlow). 🎯 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗱 𝗮𝗻 𝗔𝗜-𝗽𝗼𝘄𝗲𝗿𝗲𝗱 𝗱𝗲𝗻𝘁𝗮𝗹 𝗶𝗻𝘀𝘂𝗿𝗮𝗻𝗰𝗲 𝘃𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺 that reduced verification time by 10 minutes per transaction, utilizing OpenAI Whisper, Coqui TTS, Speech Recognition, and Gemini LLM for speech processing, with FAISS-based accent correction and AWS/PostgreSQL backend deployment. ✨ 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝗮 𝟵𝟱% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗶𝗻 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗻𝗴 𝗽𝗿𝗼𝗽𝗲𝗿𝘁𝘆 𝗽𝗿𝗶𝗰𝗲𝘀 with a margin of error within 5%, enhancing property recommendations tailored to customer preferences using K-Nearest Neighbors (KNN) and collaborative filtering algorithms, using an automated end-to-end ML pipeline. ★ 𝗜𝗺𝗽𝗿𝗼𝘃𝗲𝗱 𝘁𝗲𝗿𝗺 𝗱𝗲𝗽𝗼𝘀𝗶𝘁 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 𝗿𝗮𝘁𝗲𝘀 𝗯𝘆 𝟲𝟲% 𝗮𝗻𝗱 𝗿𝗲𝗱𝘂𝗰𝗲𝗱 𝗰𝗮𝗺𝗽𝗮𝗶𝗴𝗻 𝗰𝗼𝘀𝘁𝘀 𝗯𝘆 𝗼𝘃𝗲𝗿 𝟱𝟬% by leveraging customer segmentation and advanced machine learning models (Logistic Regression, Random Forest, K-Neighbors, XGBoost, Voting Classifier, Neural Networks). Achieved a 77% recall rate and a balanced F1 score of 0.662, significantly enhancing customer value and business profitability. 🔶 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗲𝗱 𝗿𝗲𝘁𝘂𝗿𝗻 𝗼𝗻 𝗶𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁 𝗯𝘆 𝟯𝟴% 𝗯𝘆 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗰𝗮𝗹𝗹𝘆 𝘀𝗲𝗹𝗲𝗰𝘁𝗶𝗻𝗴 𝗡𝗔𝗦𝗗𝗔𝗤-𝟭𝟬𝟬 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀 𝘂𝘀𝗶𝗻𝗴 𝟯𝟬 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗺𝗲𝘁𝗿𝗶𝗰𝘀, including risk factors such as Beneish M-Score and Altman Z-Score, alongside financial indicators like current ratio, debt to equity, and EPS growth. This is implemented in an interactive dashboard hosted on the cloud, allowing users to dynamically explore and analyze the data.

★ 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗱 𝗣𝘆𝘁𝗵𝗼𝗻 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 (Pandas, NumPy, Scikit-learn, SQLAlchemy) 𝗿𝗲𝗽𝗹𝗮𝗰𝗶𝗻𝗴 𝗳𝘂𝗹𝗹-𝘁𝗶𝗺𝗲 𝗗𝗮𝘁𝗮 𝗔𝗻𝗮𝗹𝘆𝘀𝘁 𝗿𝗼𝗹𝗲 by integrating 𝗔𝗪𝗦 𝗥𝗲𝗱𝘀𝗵𝗶𝗳𝘁 extraction and 𝗔𝗪𝗦 𝗤𝘂𝗶𝗰𝗸𝗦𝗶𝗴𝗵𝘁 visualizations, reducing analysis time from 𝟰𝟬𝗵𝗿𝘀 𝘁𝗼 𝟮𝗵𝗿𝘀 𝘄𝗲𝗲𝗸𝗹𝘆 with 𝗦𝟴𝟱𝗞 𝗮𝗻𝗻𝘂𝗮𝗹 𝘀𝗮𝘃𝗶𝗻𝗴𝘀 ➤ 𝗖𝘂𝘁 𝗲𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁𝗮𝗹 𝗰𝗼𝘀𝘁𝘀 𝗮𝗻𝗱 𝘁𝗶𝗺𝗲 𝗯𝘆 𝗼𝘃𝗲𝗿 𝟰𝟬% by detecting early battery cycling failures through advanced statistical modeling. Combined key performance indicators (KPIs) such as resistance, polarization, and discharge energy to proactively identify issues, streamlining the battery testing process. ✨ 𝗘𝗻𝗵𝗮𝗻𝗰𝗲𝗱 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻𝘀 𝗼𝗳 𝗘𝗻𝗱 𝗼𝗳 𝗟𝗶𝗳𝗲 𝗳𝗼𝗿 𝗯𝗮𝘁𝘁𝗲𝗿𝘆 𝗽𝗮𝗰𝗸𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝘃𝗮𝗿𝗶𝗼𝘂𝘀 𝘁𝗲𝗺𝗽𝗲𝗿𝗮𝘁𝘂𝗿𝗲𝘀 through innovative feature engineering. This included substituting individual cell resistance for pack resistance and calculating entropy based on the first principle of thermodynamics. ♦ 𝗗𝗲𝘀𝗶𝗴𝗻𝗲𝗱 𝘀𝘁𝗮𝘁𝗶𝘀𝘁𝗶𝗰𝗮𝗹 𝗺𝗼𝗱𝗲𝗹𝘀, 𝗔/𝗕/n 𝘁𝗲𝘀𝘁𝘀, permutation test, bootstrapping 𝗮𝗻𝗱 𝗰𝗼𝗻𝗱𝘂𝗰𝘁𝗲𝗱 𝗵𝘆𝗽𝗼𝘁𝗵𝗲𝘀𝗶𝘀 𝘁𝗲𝘀𝘁𝗶𝗻𝗴 to evaluate whether our charging algorithm meets customer requirements and outperforms the control CCCV algorithm. ★ 𝗜𝗻𝗶𝘁𝗶𝗮𝘁𝗲𝗱 𝗮𝗻𝗱 𝗹𝗲𝗱 𝘁𝗵𝗲 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗼𝗳 𝗮 𝗣𝘆𝘁𝗵𝗼𝗻 𝗽𝗮𝗰𝗸𝗮𝗴𝗲 𝗳𝗼𝗿 𝗯𝗮𝘁𝘁𝗲𝗿𝘆 𝗰𝗲𝗹𝗹 𝗽𝗮𝗿𝗮𝗺𝗲𝘁𝗲𝗿𝗶𝘇𝗮𝘁𝗶𝗼𝗻, aimed at integrating with PyBaMM simulations, which was projected to save $1M annually. ➤ 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗱 𝗧𝗮𝗶𝗹𝗼𝗿𝗲𝗱 𝗗𝗮𝘁𝗮𝘀𝗲𝘁 𝗳𝗿𝗼𝗺 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗕𝗮𝘁𝘁𝗲𝗿𝘆 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗘𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁𝘀 𝗳𝗼𝗿 𝗤𝗟𝗼𝗥𝗔 𝗣𝗘𝗙𝗧 𝗙𝗶𝗻𝗲-𝗧𝘂𝗻𝗶𝗻𝗴 𝗼𝗳 𝗺𝗲𝘁𝗮-𝗹𝗹𝗮𝗺𝗮/𝗟𝗹𝗮𝗺𝗮-𝟯.𝟭-𝟴𝗕-𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁

★ 𝗦𝗽𝗲𝗮𝗿𝗵𝗲𝗮𝗱𝗲𝗱 𝗮 𝗴𝗿𝗼𝘂𝗻𝗱𝗯𝗿𝗲𝗮𝗸𝗶𝗻𝗴 𝗮𝗱𝘃𝗮𝗻𝗰𝗲𝗺𝗲𝗻𝘁 𝗶𝗻 𝘁𝗵𝗲 𝗪𝗛𝗛 𝗺𝗼𝗱𝗲𝗹, 𝗮 𝗽𝗶𝘃𝗼𝘁𝗮𝗹 𝘄𝗼𝗿𝗸 𝗶𝗻 𝗡𝗯₃𝗦𝗻 𝗶𝗻 𝘁𝗵𝗲 𝟮𝟭𝘀𝘁 𝗰𝗲𝗻𝘁𝘂𝗿𝘆. Confirmed the refinement of 𝛼 and 𝜆ₛₒ to zero by modeling data on hundreds of Nb₃Sn samples with varying parameters 𝛼 and 𝜆ₛₒ. This eliminates the need for 31 T measurements to accurately determine Hc2(0 K), providing a cost-efficient, consistent, and reliable approach. ➤ 𝗘𝗺𝗽𝗹𝗼𝘆𝗲𝗱 𝗮𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗘𝗧𝗟 𝘁𝗲𝗰𝗵𝗻𝗶𝗾𝘂𝗲𝘀 𝗼𝗻 𝗲𝘅𝘁𝗲𝗻𝘀𝗶𝘃𝗲 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱/𝘂𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱 𝗱𝗮𝘁𝗮𝘀𝗲𝘁𝘀 using SQL and Python’s Big Data packages, including NumPy, Pandas, scikit-learn, Keras, TensorFlow, PyTorch, Seaborn, and Matplotlib. Extracted key patterns, insights, and predictions. ✨ 𝗨𝘁𝗶𝗹𝗶𝘇𝗲𝗱 𝗮 𝗿𝗮𝗻𝗴𝗲 𝗼𝗳 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗮𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺𝘀, including Linear Regression, Random Forest Regression, K-NN Regression, Gradient Boosting Regression, LSTM Networks, ARIMA, and Support Vector Regression. Modeled intricate relationships between LTS electromagnetic properties and numerous experimental factors, forecasting time series trends. ♦ 𝗜𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗲𝗱 𝗽𝗿𝗼𝗰𝗲𝘀𝘀 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝗺𝗲𝗻𝘁𝘀 𝗶𝗻 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 through the successful application of lean and Six Sigma techniques, including DOE (design of experiments) and SPC (statistical process control). Achieved cost reduction and enhanced efficiency. ➤ 𝗣𝘂𝗯𝗹𝗶𝘀𝗵𝗲𝗱 𝗵𝗶𝗴𝗵-𝗾𝘂𝗮𝗹𝗶𝘁𝘆 𝗮𝗰𝗮𝗱𝗲𝗺𝗶𝗰 𝗷𝗼𝘂𝗿𝗻𝗮𝗹𝘀 𝗳𝗼𝗿 𝗽𝗿𝗲𝘀𝘁𝗶𝗴𝗶𝗼𝘂𝘀 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗽𝘂𝗯𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 and delivered comprehensive reports for internal and external stakeholders. ✔ 𝗥𝗲𝗰𝗼𝗴𝗻𝗶𝘇𝗲𝗱 𝗮𝘀 𝗮𝗻 𝗼𝘂𝘁𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 𝗺𝗲𝗻𝘁𝗼𝗿 𝗳𝗼𝗿 𝗲𝗳𝗳𝗲𝗰𝘁𝗶𝘃𝗲𝗹𝘆 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗻𝗲𝘄 𝗵𝗶𝗿𝗲𝘀 in data analytics and manufacturing process improvement.

★ 𝗖𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗲𝗱 𝘄𝗶𝘁𝗵 𝗮𝗻 𝗮𝗰𝗮𝗱𝗲𝗺𝗶𝗰 𝘀𝘂𝗽𝗲𝗿𝘃𝗶𝘀𝗼𝗿 𝘁𝗼 𝗱𝗲𝘃𝗲𝗹𝗼𝗽 𝗮𝗻𝗱 𝗶𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝘃𝗲 𝘁𝗲𝗮𝗰𝗵𝗶𝗻𝗴 𝗺𝗲𝘁𝗵𝗼𝗱𝘀 𝗮𝗻𝗱 𝗰𝘂𝗿𝗿𝗶𝗰𝘂𝗹𝗮𝗿 𝘃𝗶𝗲𝘄𝗽𝗼𝗶𝗻𝘁𝘀. ➤ 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗱 𝗮𝗻𝗱 𝗲𝘅𝗲𝗰𝘂𝘁𝗲𝗱 𝗲𝗻𝗴𝗮𝗴𝗶𝗻𝗴 𝗮𝗻𝗱 𝘁𝗵𝗼𝘂𝗴𝗵𝘁-𝗽𝗿𝗼𝘃𝗼𝗸𝗶𝗻𝗴 𝗲𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁𝘀, utilizing relatable real-world examples to enhance the learning experience and increase student engagement. ✔ 𝗜𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗲𝗱 𝗱𝗮𝘁𝗮 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝘁𝗼𝗼𝗹𝘀, including Tableau, Power BI, Excel, and SciPy, to guide students in analyzing and interpreting data, fostering critical thinking and understanding of real-world applications. ✨ 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗱 𝗮𝗻𝗱 𝗮𝗱𝗺𝗶𝗻𝗶𝘀𝘁𝗲𝗿𝗲𝗱 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁𝘀, including tests and quizzes, to evaluate student understanding and progress. ♦ 𝗣𝗿𝗼𝘃𝗶𝗱𝗲𝗱 𝗰𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝘃𝗲 𝗳𝗲𝗲𝗱𝗯𝗮𝗰𝗸 𝗮𝗻𝗱 𝗴𝗿𝗮𝗱𝗲𝗱 𝗽𝗮𝗽𝗲𝗿𝘀, regularly publishing grades to ensure transparency and accountability. ★ 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝗵𝗶𝗴𝗵 𝘀𝘁𝘂𝗱𝗲𝗻𝘁 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗮𝗻𝗱 𝗲𝗻𝗴𝗮𝗴𝗲𝗺𝗲𝗻𝘁 through an interactive and participatory teaching approach, offering dedicated office hours for student feedback and support.

𝗚𝗿𝗮𝗱𝘂𝗮𝘁𝗲 𝗦𝘁𝘂𝗱𝗶𝗲𝘀 𝗶𝗻 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 & 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 (𝟮𝟰 𝗖𝗿𝗲𝗱𝗶𝘁 𝗛𝗼𝘂𝗿𝘀) - Machine Learning and Computer Vision - Programming for Chemist/Biochemist (Python/R/HTML/CSS) - Bioinformatics (Python/R) - Computational Scientific Research - Special Topics CS NM Python Programming - Numerical Solutions (MATLAB, C++, Python) - Statistical Mechanics I & II 💻 𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗦𝗸𝗶𝗹𝗹𝘀 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀: Python, SQL, R, MATLAB, C++ 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 & 𝗔𝗜: - 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀: Pandas, NumPy, SciPy, SpaCy, NLTK, PySpark, Scikit-learn, TensorFlow, PyTorch, JAX - 𝗠𝗼𝗱𝗲𝗹𝘀: GLM, Random Forest, XGBoost, LightGBM, CatBoost, SVM, CNN, RNN, LSTM, GRU, MLP - 𝗨𝗻𝘀𝘂𝗽𝗲𝗿𝘃𝗶𝘀𝗲𝗱 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴: KMeans, DBSCAN, Hierarchical Clustering, GMM 𝗗𝗮𝘁𝗮 𝗩𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻: Matplotlib, Seaborn, Plotly, PowerBI, Tableau 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 & 𝗖𝗹𝗼𝘂𝗱: JavaScript, HTML/CSS, AWS, Jenkins

Skills: Design of Experiments (DoE), Data Analysis and Visualizations, Programming (Python, SQL, R, MATLAB), Machine Learning, Mentorship, Leadership, Cross-functional collaboration, Technical writing and publications, Communication, Presentation, Lean and Six Sigma Methodologies, Microsoft Office Products