Experience
- Wrote 3 backend tools using Python and exposed FastAPI endpoints that implemented OCR, page orientation detection, and text blur detection using OpenCV libraries with <5% margin error
- Implemented each tool into an agentic AI workflow to automate Quality control for Wells Fargo’s loan application process and successfully reduce Loan Analyst review time by 80%
- Built and deployed a GPU-accelerated live speech-to-text endpoint on CodeEngine using IBM’s 8B-parameter Speech model
- Integrated the speech-to-text endpoint into an agentic AI workflow with parallel processing for summarization, sentiment analysis, and recommended actions—reducing support delivery time by 50%
- Built an Agentic RAG system using Langflow and AstraDB, vectorizing 100+ 10-K filings; containerized with Docker, deployed to IBM CodeEngine, and integrated as a custom tool in Watsonx Orchestrate
- Led a technical team of interns to develop a Full-Stack internal marketplace application, which led to buy-in from C-Suite Executives
PythonFastAPIOpenCVIBM CodeEngineDockerLangflowAstraDBWatsonxReact
- Creating a software to streamline the NDA contract negotiation process with a FastAPI Python backend that currently has 3k+ lines of code, and a React JavaScript Frontend with 5 pages currently deployed
- Integrated Supabase (PostgreSQL + storage), Google Firebase Auth, GitHub, Vercel, and Render into a seamless CI/CD workflow, automating database migrations and deployments across those 5 platforms to reduce deployment time by 40%
- Implementing document editing features including redline tracking, version control, and automated email notifications
PythonFastAPIReactJavaScriptSupabasePostgreSQLFirebase AuthVercelRenderCI/CD
- Performed data cleaning, model tuning, and prompt engineering in Python Notebooks and compared outputs from 9 Watsonx.ai large language models in a RAG use case for Honda vehicle design
- Utilized IBM prompt lab to generate 20 examples of input and output data in Python Dictionary format to be used for RAG metric testing in developing a watsonx.gov dashboard for PepsiCo’s Gen AI models
PythonWatsonx.aiRAGJupyter NotebooksPrompt EngineeringLLMs
- Researched the applications of AI and ML in IBM’s broad range of 7+ cybersecurity products
- Developed 7 innovative use cases, using my knowledge of AI & ML, that I presented to IBM Cybersecurity sales managers
AI/MLCybersecurityIBM ProductsTechnical Sales
- Operated my own shoe re-selling business and brought $20k in revenue
- Programmed a software in C# that used a web-driver to acquire inventory from the Footlocker website
C#Web ScrapingE-commerceBusiness Operations
- Generated recurring yearly sales revenue of approximately $71,000 on behalf of AT&T
- Negotiated benefits of AT&T services to over 2,000 businesses and corporations through cold calling
SalesAccount ManagementAT&T ServicesBusiness Development
Education
- Focus in Artificial Intelligence
- Coursework: Applied Machine Learning, Deep Learning for Healthcare, Natural Language Processing, Computational Photography, Software Engineering, Practical Statistical Learning
- Honors/Scholarships: Academic Excellence Scholar, JSOM Freshman Excellence Scholar
- Application of Statistics, Data Science, and Machine Learning techniques through Python programming
- Completed 9 projects based on real-world business applications
Skills
Programming:
PythonJavaJavaScriptTypeScriptSQLHTMLCSSXMLC#
Frameworks / Tools:
FastAPIReact.jsGitVercelRenderSupabaseFirebaseOpenCVLangFlowLangChainPandasNumPyMatplotlibSciPyTableauMicrosoft OfficeBlender
Concepts:
Machine LearningGenerative AIData ScienceBackend/Frontend DevelopmentCI/CDApplication Development
Projects
PythonNumPyOpenCVSciPyBlender
- •Implemented texture synthesis pipeline via dynamic programming-based image quilting, generating realistic imagery, and processing 8+ texture samples across 2,015 lines of implementation code
- •Developed Gradient Domain Processing using Poisson equation solvers and Laplacian pyramids across 1,925 lines of python notebook code for seamless object insertion and blending
- •Built HDR reconstruction and Object rendering system by merging 6+ exposure sequences with equirectangular environment mapping, integrating Blender's Cycles engine for GPU-accelerated ray-traced rendering
JavaAndroidXMLGoogle Maps APIWeather APILLM APIGradle
- •Built an Android app in a team of 5 Software Engineers, enabling users to add global locations, view them on Google Maps, fetch real-time weather data via API, and display insights using an LLM API
- •Developed 10 backend Java files and 10 frontend XML files; implemented both manual and LLM-generated Gradle testing
PythonTensorFlowOpenCVDeep LearningCNNVGG16Data Augmentation
- •Constructed an ML model on a dataset of 1000+ images that identifies Brain MRI Scans with or without a tumor
- •Utilized Data Augmentation then a 16-layer Convolutional Neural Network with Dropout, Pooling, and other Deep Learning concepts to achieve accuracy of 82%
- •Leveraged the open-source VGG16 CNN model, which achieved 95% accuracy on the validation set
PythonNLPTokenizationVectorizationMachine LearningSentiment Analysis
- •Applied Python, Tokenization, Vectorization, and other Natural Language Processing techniques to identify positive, neutral, and negative sentiments toward U.S. Airlines based on 15 parameters of over 14,000 customer tweets
- •Developed an ML model that helps U.S. Airlines identify customer sentiment and achieved accuracy of 80%