Shelton Bumhe | Backend Systems & Software Engineering

01 / ABOUT

I build software that makes operational work easier to see.

My best work sits between backend engineering, product thinking, and real-world operations.

Core approach

I start with the workflow: who needs visibility, what information is missing, where work slows down, and which system boundary should own the fix. Then I choose the technology that fits.

01Understand the real user and operational constraint.

02Model the workflow clearly across services, events, and data.

03Build an interface that makes system state useful.

02 / EXPERIENCE

Engineering shaped by the operation.

Experience across enterprise backend work, operational data workflows, teaching, and leadership development.

Current FISERV

Enterprise systems

Fiserv

Software Engineering Intern

Backend · Events · Data

Working on an enterprise file-tracking platform built to solve visibility gaps across large-scale file-processing workflows. I contribute to backend services, event modeling, business-rule validation, SQL data modeling, and searchable APIs that help teams understand file movement, status, failures, and relationships across internal systems.

Contributed to backend services for a platform designed to process and monitor up to 1M file events per day.
Helped model file movement, processing status, failures, and relationships using structured events and SQL-backed tracking records.
Used Java and Spring Boot to support service-layer logic, API behavior, and business-rule validation for searchable file workflows.
Supported event-driven architecture planning using Kafka concepts for scalable, asynchronous file updates.
Contributed to an AI-assisted development workflow for code structure, documentation, and alignment with engineering standards.
Connected technical design to faster issue detection and potential savings of up to $100K per month in manual investigation and processing expenses.

JavaSpring BootKafkaSQLREST APIsAI-assisted development

2022—202302

Dominion College

Software Developer & Math Tutor

Built a Python and SQL tracking workflow to organize student records, tutoring activity, and progress reporting when manual processes made it difficult to see who needed help and where students were improving.

PythonSQLReporting workflows

PROFESSIONAL DEVELOPMENT03

Management Leadership for Tomorrow

MLT Career Prep Fellow

Developing communication, leadership, technical interview preparation, networking, and product thinking in a high-accountability environment. That founder-minded perspective also informs Grand Pilot, my concept for connecting small businesses with mentors and experienced founders.

LeadershipProduct thinkingTechnical interviews

03 / SELECTED SYSTEMS

Systems Built Around Real Problems

I build with the user problem first, then choose the technology that fits.

Fiserv Enterprise File Tracking Platform

Software Engineering Intern · Backend services · Event modeling · Search APIs

File EventBusiness RulesSQL HistorySearch APIDashboard

1Mfile events / day

$100Kpotential savings / month

CID → statussearch across file, job, client, and stage

Problem

Large-scale file workflows become difficult to monitor when files move across multiple internal systems without one tracking view. Teams need to see state, failures, relationships, and downstream processing.

What I worked on

I contributed to the design of backend services that capture file events, apply business rules, store processing history, and expose searchable tracking data for operational visibility.

How the system works

File updates become consistent records, persist in a relational database, and flow through APIs so teams can search by CID, file name, job number, client, status, and processing stage.

Technologies used

Java and Spring Boot support service logic, API endpoints, validation, and workflow handling. Kafka concepts support asynchronous event planning. SQL models history, reconciliation details, relationships, and reporting data. An internal AI workflow supports code structure and documentation.

Scale / impact

Designed around up to 1M file events per day. The platform aims to reduce manual investigation and help teams detect missing, delayed, failed, or misrouted files faster—with potential operational savings of up to $100K per month.

JavaSpring BootKafkaSQLAPIsEnterprise SystemsAI-Assisted Development

Grand Pilot

Founder & Small Business Connection Platform

Problem: Many small business owners need practical guidance but do not always have access to experienced founders, mentors, or trusted networks.
What I worked on: Defined the product concept, user problem, platform flow, and mentorship connection model.
How it works: Owners discover founders and mentors, request conversations, and receive guidance on growth, operations, and strategy.
Methods used: Product thinking and user research shape the discovery, matching, request, and conversation flows before implementation.
Expected value: Makes founder knowledge and practical mentorship more accessible to small businesses.

Product ThinkingEntrepreneurshipMentorship PlatformUser Research

Dominion Student Tracking

A clearer workflow for student progress and tutoring activity

Problem: Manual tracking made it harder to review progress, attendance, tutoring notes, and performance updates.
What I worked on: Built the Python and SQL-based tracking workflow for student records, tutoring activity, and weekly progress reports.
How it works: Python organizes updates into a consistent reporting flow while SQL stores structured student and tutoring records for review.
Technologies used: Python handled record processing and report workflows; SQL made progress and tutoring data consistent and queryable.
Impact: Reduced manual tracking effort and made student progress easier for staff to review.

PythonSQLReportingEducation Technology

AY Logistics Drone Inspection Platform

A visual workflow for safer, more consistent vehicle inspections

Problem: Truck and trailer inspections can be slow, repetitive, and inconsistent when completed only through manual walk-arounds.
What I worked on: Designed a drone-based inspection workflow with React dashboard and Python/FastAPI backend planning.
How it works: Operators manage inspections, review drone-captured images, tag issues, track vehicle history, and follow up on maintenance.
Technologies planned: React structures the operations dashboard; Python and FastAPI define the planned inspection and image-data services; computer vision remains an exploration area.
Expected value: Could reduce inspection time, improve documentation, and give logistics teams better maintenance visibility.

React DashboardPythonFastAPIComputer VisionOperations

Retail Intelligence / E-commerce Platform

A commerce system designed beyond basic CRUD

Problem: Commerce systems need reliable transactions, fast catalog reads, protected checkout paths, and a clear path to deployment.
What I worked on: Built the user workflows, API and business-logic layers, relational data model, caching strategy, integrations, and deployment workflow.
How it works: React handles catalog, cart, authentication, and checkout interactions. ASP.NET Core owns business logic and APIs, while PostgreSQL persists products, users, and orders.
Technologies used: Redis reduces repeated catalog reads; Stripe isolates payment processing; Cloudinary manages product media; GitHub Actions and Azure support delivery.
Impact: Demonstrates production-minded decisions around caching, protected routes, horizontal scaling, and release automation.

C#ASP.NET CoreReactPostgreSQLRedisAzure

GitHub ↗

NFL QB Touchdown Predictor

Prediction paired with explanation and a usable interface

Problem: A prediction is less useful when users cannot understand the data behind it or the features influencing the result.
What I worked on: Handled data preparation, feature engineering, model training, evaluation, explainability, and the Streamlit product interface.
How it works: Historical game data becomes model features, XGBoost predicts touchdown outcomes, and SHAP shows which inputs pushed each result.
Technologies used: Pandas supports cleaning and feature preparation; XGBoost performs classification; SHAP adds model explanation; Streamlit exposes the workflow to users.
Demonstrates: How to connect an ML pipeline to an interpretable, user-facing product instead of stopping at a notebook.

Reported model accuracy88%trained on 10,000+ plays

PythonPandasXGBoostSHAPStreamlit

GitHub ↗

04 / CAPABILITIES

Tools organized by the work they do.

I care about systems that are useful, not just impressive.

{ }

Backend

Building APIs, service layers, business logic, and system workflows.

Java · Spring Boot · C# · ASP.NET Core · REST APIs · GraphQL

Databases

Designing schemas, relationships, indexes, and query flows for application data.

SQL Server · PostgreSQL · MySQL · Oracle · JPA · JDBC

⇢

Event-driven systems

Modeling workflows as events and processing updates asynchronously.

Kafka · Event modeling · Workflow state

Frontend

Building clear interfaces that expose system capabilities to users.

React · JavaScript · TypeScript

∞

Cloud & DevOps

Packaging, validating, and delivering software through repeatable workflows.

Docker · GitHub Actions · GitLab · Azure

AI & Data

Turning data into models, explanations, and developer-assisted workflows.

Python · Pandas · scikit-learn · Machine learning · AI-assisted tools

05 / PORTFOLIO ASSISTANT

Ask Shelton AI

A focused guide to my experience, projects, technical strengths, architecture decisions, and career goals. It answers from a structured portfolio knowledge base.

01 Recruiter-friendly summaries
02 System and architecture context
03 Project-specific technical detail

Structured demo · No personal data collected

06 / CONTACT

Let’s build something useful.

I’m interested in software engineering opportunities where backend systems, real user problems, and strong engineering judgment matter.

Emailsbumhe2@huskers.unl.edu LinkedInConnect with me ↗ GitHubView my code ↗