Home

Diwita Banerjee - AI / ML Enginner

[email protected]

Location: Fairfax, Virginia, USA

Relocation:

Visa:

Resume file: Diwita_Banerjee_AI_ML_1756489196207.pdf Please check the file(s) for viruses. Files are checked manually and then made available for download.

Diwita Banerjee AI/ML Engineer +17035856489 | [email protected] | LinkedIn | GitHub | Open to Relocate PROFESSIONAL SUMMARY AI/ML Engineer with 5+ years of experience designing and deploying production-grade machine learning solutions in financial services, healthcare, and enterprise domains. Skilled in Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Natural Language Processing (NLP), and MLOps. Experienced in fine-tuning Transformer models, building scalable APIs, and deploying AI systems on AWS, GCP, and Azure. Delivered LLM-powered pipelines that improved document retrieval accuracy by 30% and reduced inference latency by 40%. Automated ML workflows with Airflow, MLflow, Jenkins, and Docker, cutting deployment cycles by 50%. Strong background in regulatory-compliant AI, feature engineering, and real-time analytics. Adept at collaborating with cross-functional teams to drive business impact through AI adoption. TECHNICAL SKILLS Languages & IDEs: Python, SQL, Shell Scripting, MATLAB, Jupyter Notebook, VS Code, Google Colab, SSMS Frameworks & Libraries: Pandas, NumPy, Scikit-learn, TensorFlow, Keras, PyTorch, XGBoost, Matplotlib, Seaborn, NLTK, spaCy, Sentence-BERT, Hugging Face Transformers, LangChain Machine Learning: Logistic Regression, Linear Regression, Decision Trees, Random Forests, SVM, Naive Bayes, A/B Testing, Model Evaluation, Feature Engineering Deep Learning & NLP/LLMs: CNN, RNN, LSTM, BERT, RoBERTa, GPT-3/4, LLMs, Prompt Engineering, RAG pipelines, Fine-tuning (LoRA/PEFT) MLOps & Deployment: Docker, Kubernetes (EKS), FastAPI, Flask, MLflow, Jenkins, Terraform, Airflow, PySpark, Model Drift Detection, Usage Metrics Dashboards Cloud & Visualisation: AWS (Lambda, EC2, S3, RDS, SageMaker, SQS, SNS, CodeDeploy, CloudWatch, API Gateway), GCP, Azure, Tableau, Power BI Databases & Tools: PostgreSQL, MySQL, MongoDB, Cassandra, Redis, Neo4j, SQL Server, SQLAlchemy Statistical Techniques: Hypothesis Testing, Data Modeling, Data Visualization, Experiment Design, A/B Testing Collaboration Tools: Agile Development, Requirements Gathering, Stakeholder Engagement, AI Insights Visualization PROFESSIONAL EXPERIENCE Freddie Mac | AI/ML Engineer Jun 2024 Present Designed and implemented LLM-powered RAG pipelines using FAISS and Hugging Face, increasing financial document retrieval accuracy by 30%. Built and deployed FastAPI-based microservices on AWS EKS and Lambda, reducing model inference response time by 40% and improving scalability. Fine-tuned Transformer models, including BERT and RoBERTa, on compliance-related datasets, improving contextual relevance by 25%. Automated end-to-end ML lifecycle management using Airflow, MLflow, Jenkins, and Docker, shortening deployment cycles by 50%. Developed monitoring dashboards with PySpark and AWS CloudWatch for drift detection and performance analysis of deployed models. Collaborated with quantitative analysts and compliance officers to deliver regulator-compliant AI solutions with transparent auditability. George Mason University | AI Research Assistant March 2024 Nov 2024 Built fairness and bias evaluation pipelines for multimodal AI models, including Stable Diffusion, LLaVA, and InstructBLIP, uncovering demographic disparities. Developed a medical VQA system using MIMIC-IV EHR data that reduced clinician query resolution time by 25% in controlled trials. Optimised large-scale training and inference workflows on the Hopper HPC cluster, reducing runtime by 40%. Conducted federated multimodal experiments for privacy-preserving healthcare AI and improved model generalization in clinical settings. Benchmarked multiple Vision-Language Models (VLMs) and contributed results to academic publications. Co-authored ongoing research papers in responsible and federated healthcare AI. Cognizant Technology Solutions | Machine Learning Developer, Salesforce CRM Nov 2021 Jul 2023 Built predictive lead/case/renewal scoring models with Einstein Prediction Builder and automated nightly training and scoring via Apex batch and queueable jobs. Delivered prescriptive recommendations using Einstein Discovery and Next Best Action, boosting SLA compliance and agent productivity. Develop CRM Analytics datasets and dashboards to monitor win-rate lift, adoption, precision/recall, and model drift, with alerts for threshold deviations. Integrated external ML endpoints through MuleSoft and Named Credentials with retry, timeout, and audit logging, ensuring secure and reliable data exchange. Implemented event-driven and batch pipelines with Platform Events and REST APIs for near real-time updates and high-volume processing. Automated CI/CD pipelines with Git and Copado, added unit/integration tests for high coverage, and authored runbooks to streamline monitoring and rollback. Recognized with the Working as One (2023) and Always Striving & Never Settling (2022) awards. Larsen & Toubro Infotech | AI Developer Full Stack Applications Jun 2020 Oct 2021 Engineered AI-driven predictive analytics modules using Spring Boot, Python, and Oracle SQL, reducing backend response latency by 39%. Developed intelligent data-driven UI components with React.js, improving usability and long-term scalability. Automated defect root cause analysis workflows, reducing issue resolution time by 30%. Designed APIs for scalable integration of AI models into enterprise applications. Collaborated with business stakeholders to align AI feature development with operational requirements. Integrated backend analytics with enterprise dashboards for real-time monitoring and decision-making. RESEARCH & PROJECTS MedBLIP Medical Image Captioning (GMU, Spring 2025) Fine-tuned BLIP on the ROCO dataset for medical imaging captioning tasks. Achieved benchmark improvements across CIDEr, SPICE, and BERTScore. Compared performance with BLIP-2, InstructBLIP, and Gemini models. CloudMart Multicloud AI E-Commerce Platform (Bootcamp, Mar 2025) Designed and deployed a GPT-powered shopping assistant across AWS, GCP, and Azure using Terraform, Kubernetes, and Docker. Implemented real-time product discovery, intelligent Q&A, and analytics pipelines. Built cross-cloud ETL-style AI pipelines with Git-based CI/CD, improving scalability and observability. News Summarisation & Simplification (GMU NLP Project, Fall 2024) Built an LSTM-based summarizer and T5 simplification model using Hugging Face Transformers and TensorFlow. Trained on the WikiAuto dataset and achieved BLEU 0.22 and Flesch-Kincaid readability 8.31. Delivered an end-to-end NLP pipeline to improve accessibility and comprehension of complex news content. Movie Recommendation (IJARESM, 2022) Developed a BERT-based semantic search and recommendation engine on the CMU Movie Summaries dataset. Enhanced semantic retrieval and recommendations with a 15% improvement in accuracy. Published results in the IJARESM research journal. EDUCATION Master of Science in Computer Science (Machine Learning Concentration) George Mason University, Fairfax, VA Coursework: Algorithms, AI, Data Mining, Machine Learning, Deep Learning, Advanced NLP Aug 2023 May 2025 Honours & Activities: Led and Won 2nd Position in GDG-GMU Hackathon; research on RCRS for Prompt Injection Mitigation Keywords: continuous integration continuous deployment artificial intelligence machine learning user interface javascript business intelligence sthree Colorado Virginia