AI Engineer: ML, NLP, MLOps, Cloud Deployments

SRIYA KAKKERA

Michigan, USA 947-***-**** ************@*****.*** LinkedIn

SUMMARY

Results-driven AI Engineer with 3 years of experience designing, building, and deploying machine learning and AI-driven solutions across IT services and financial services domains. Strong expertise in applied machine learning, deep learning, NLP, MLOps, and cloud-based AI platforms. Proven ability to translate complex business problems into scalable data science solutions that drive measurable business impact in enterprise environments.

SKILLS

Machine Learning & AI:Supervised Learning, Unsupervised Learning, Deep Learning, CNN, RNN, LSTM, Transformer Models, Time Series Forecasting, Recommendation Systems

Computer Vision: Object Detection (YOLOv4/v5), Object Tracking (DeepSORT), Image Preprocessing, OpenCV, Image Classification Natural Language Processing: Text Preprocessing, Tokenization, Embeddings, Sentiment Analysis, Transformer-based NLP Programming & Libraries: Python, PySpark, NumPy, Pandas, Scikit-learn, TensorFlow, PyTorch, XGBoost Data Engineering & Databases: SQL, Snowflake, Databricks, Apache Spark, ETL, Data Warehousing MLOps & Deployment: AWS SageMaker, Docker, Kubernetes, FastAPI, CI/CD, Model Monitoring, Versioning Cloud Platforms: AWS (S3, Lambda, EC2, SageMaker)

Visualization & Reporting: Tableau, Power BI, Matplotlib, Seaborn Tools & Collaboration: Git, JIRA, Agile/Scrum, Confluence PROFESSIONAL EXPERIENCE

Capital One, Michigan, USA Aug 2024 -

Present

AI Engineer

● Designed and deployed a real-time fraud scoring system using Apache Kafka, processing ~10 million transaction events per day with sub-second latency, improving fraud detection coverage across payment channels.

● Built LSTM-based sequence models in PyTorch to analyze transaction timelines, reducing false positives by ~18% compared to existing rule-based systems while maintaining detection accuracy.

● Implemented automated model pipelines with AWS SageMaker Pipelines, reducing model release cycle time by ~30% through standardized training, validation, and version control for production models.

● Delivered model interpretability using SHAP, generating feature-level explanations for 100% of high-risk transactions, enabling compliance teams to review and approve model-driven decisions faster.

● Developed production monitoring using Evidently AI, detecting data drift and performance degradation within 24 hours, which helped prevent potential revenue losses from model drift.

● Integrated fraud risk predictions into enterprise authorization systems through AWS Lambda inference services, supporting real-time decisions across web, mobile, and card networks and improving response consistency. Cognizant, India Jan 2022 – Jul

2023

Data Scientist Associate

● Analyzed over 5 million customer records across telecom and retail datasets to identify churn patterns and define retention strategies, improving stakeholder clarity on key risk drivers.

● Cleaned and engineered features using Python (Pandas, NumPy), creating 60+ derived behavioral and transactional variables, which increased model accuracy and reduced feature noise.

● Built XGBoost churn prediction models achieving 82% accuracy and AUC of 0.88, enabling targeted retention campaigns for the top 10% high-risk customers.

● Developed a collaborative filtering recommendation system using matrix factorization, improving product recommendation relevance by ~20% based on offline evaluation metrics.

● Built demand forecasting models with Facebook Prophet, achieving 15% reduction in forecasting error (MAPE) and supporting operational teams in capacity planning and inventory allocation.

● Deployed ML models through FastAPI REST endpoints, enabling real-time predictions for client applications and reducing manual decision time by 35%.

PROJECTS

Retrieval-Augmented Generation (RAG) API - Local LLM System, Python, FastAPI, Ollama, ChromaDB, Swagger UI

● Built a local Retrieval-Augmented Generation (RAG) system using Ollama (TinyLlama) and ChromaDB to retrieve relevant context and generate accurate responses. Implemented REST APIs with FastAPI, validated endpoints with Swagger UI, and enabled dynamic knowledge-base updates for consistent output.

Road Traffic Vehicle Detection & Tracking - Python, YOLOv4, DeepSORT, OpenCV, NumPy, CUDA

● Developed a real-time vehicle detection and tracking system using YOLOv4 and DeepSORT with OpenCV, enabling automated traffic flow analysis and congestion monitoring through frame-by-frame object association and movement analytics. EDUCATION

Master of Science in Data Science - Wayne State University, Detroit, MI, USA Bachelors of Technology in Computer Science and Engineering - St. Martins Engineering College, Hyderabad, India

Contact this candidate