Viraj Sabhaya

Software Engineer | Backend & Infrastructure

I build backend services and the tools to deploy, troubleshoot, and maintain them. Healthcare systems at Altera Digital Health. Open-source work in reliability and software dependency analysis.

Dallas-Fort Worth, Texas

Viraj Sabhaya profile animation

About

Backend systems. Practical delivery.

At Altera Digital Health, I work on terminology services used by EHR applications. That includes modernizing interfaces, troubleshooting SQL Server, automating releases, and working with QA and domain specialists to preserve existing behavior while improving the system.

My open-source work explores crash recovery, retry safety, and software dependency evidence. I value clear operating constraints, reproducible tests, and human oversight when AI tools interact with enterprise systems. My research background adds experience with graph analysis and making complex data easier to inspect.

In practice

Delivery and reliability.

50%

Reduced QA server count from 14 to 7 through consolidation, reducing the number of environments to maintain.

CI/CD

Automated Web API builds and on-premises IIS delivery, including tests, versioned artifacts, SQL updates, and health checks.

30 cases

Deterministic crash/retry scenarios in EffectFence's model-based benchmark, covering unsafe schedules and safe controls.

Experience

Where I've worked.

05/2024 - Current Remote

Software Engineer at Altera Digital Health

Healthcare terminology engineering with C#/.NET, SQL Server, Python, TypeScript, PowerShell, and Azure DevOps.

  • Developed backend features for a team modernization effort spanning REST services, SOAP compatibility, and a .NET client package, with existing EHR behavior as a migration constraint.
  • Built Azure DevOps release automation for tested, versioned artifacts, on-premises IIS delivery, SQL updates, backups, and health checks; extended standalone IIS and database installation tooling.
  • Traced five-minute SQL queries on multi-million-row tables to stale statistics, and evaluated two-phase paging, indexing, full-text search, and query-plan alternatives.
  • Worked with QA on legacy/new service comparisons, clinical content defects, and deployment validation. Documented architecture and installation workflows, and helped teammates navigate the terminology domain.
  • Built a Copilot-powered VS Code prototype for terminology data-quality investigations, combining read-only SQL evidence, documentation retrieval, and response drafts for a human reviewer.
  • Authored repository context and Copilot instructions for bug-fix regression coverage and validation beyond unit tests.
09/2022 - 05/2024 UT Arlington

Software Developer at IT Lab

Co-developed the MLN Dashboard using Python, Pandas, NetworkX, and Matplotlib for multilayer graph ingestion, analysis, and visualization. Improved Louvain and Infomap workflows and added interactive comparisons for airline, LinkedIn, and IMDb datasets. This research contributed to an IEEE BigDataService publication and a VLDB demo paper.

06/2022 - 08/2022 Addison, TX

Software Engineer Intern at Securonix

Automated Cypress testing for the SNYPR SIEM platform, contributed to Kafka and Spark data-pipeline automation, and shipped QA tooling that reduced manual work and supported faster, more dependable sprint delivery.

Selected work

Systems, tools, and research.

Altera Digital Health | Professional work

Healthcare terminology platform

My contributions to a team modernization effort connect backend feature development, compatibility fixes, and repeatable delivery. The work supports terminology lookup and clinical content workflows used by EHR applications.

C#/.NET SQL Server PowerShell Azure DevOps IIS
Feature and bug-fix PRs
Implemented effective-date filtering, retired-term retrieval, and exact-code/alias search fixes, with focused regression tests to protect legacy behavior.
Service compatibility
Developed features across REST services and the .NET compatibility package, and worked on SOAP/API integration so existing clients could retain their contracts.
Deployment and installer tooling
Extended IIS and database installers with authentication options, prerequisite and HTTPS certificate checks, application-pool configuration, permission handling, and install/uninstall workflows.
Verification and handover
Compared legacy and new endpoint behavior, investigated defects through request logs and SQL, and documented architecture and installation procedures for QA and handover.
Open source | Reliability

EffectFence

A Python crash/retry and MCP side-effect conformance verifier. Explores modeled failure schedules and checks observed effects for duplication, loss, and stale retries.

Includes a 30-case deterministic benchmark and a separate live Kafka/PostgreSQL crash test between sink commit and offset commit. Reports include traces and content-addressed evidence; results are bounded by the tested schedules and configured observers.

View code and verification scope
Open source | Supply-chain security

SBOM Evidence Fusion

A Python tool that reconciles CycloneDX/SPDX documents and package-manager dependency evidence while preserving source provenance and uncertainty.

Reports dependency-graph reachability as reachable, unreachable, or unknown, with a 100-case synthetic benchmark. The analysis concerns dependency relationships, not code-level exploitability; real-world comparative validation remains future work.

View code and benchmark scope
Research | Data systems

MLN-geeWhiz

Co-developed the MLN Dashboard for multilayer network generation, analysis, and interactive visualization using Python, Pandas, NetworkX, and Matplotlib.

Co-authored related research published at IEEE BigDataService in 2023 and a VLDB demo paper in 2025, connecting graph-processing workflows with usable research tools.

View research dashboard
Hackathon | Applied AI

Load-N-Ask

Co-built a document question-answering application with Streamlit, LangChain, FAISS, and the OpenAI API, using document chunking and vector retrieval to ground responses.

Winner of the Best Use of Streamlit category at HackUTA 2023.

View project

Applied AI

AI-assisted engineering, grounded in evidence.

I use GitHub Copilot Chat to investigate unfamiliar systems, develop targeted changes, and build repeatable engineering workflows. My application work includes retrieval-augmented generation and tool-connected data-quality investigation.

Repository context and workflow design
Created project reference documents and Copilot guidance so agents can work with existing architecture, constraints, and conventions. Added bug-fix validation guidance that calls for regression coverage and checks beyond unit tests.
AI-assisted debugging and delivery
Use Copilot to trace code paths, investigate SQL and service behavior, draft fixes, and develop tests. I check the resulting diffs and behavior against logs, requirements, and legacy-system comparisons before accepting a change.
Tool-connected investigation
Built a Copilot-backed agent loop with structured tool calls, read-only SQL evidence, and local documentation search. Extended the assistant with configurable product context and editable project knowledge.
Retrieval-augmented generation
Co-built Load-N-Ask with Streamlit, LangChain, FAISS, and the OpenAI API, using document chunking and vector retrieval to ground answers in supplied material.
Verification and human control
Model output is a starting point, not proof. My workflow uses scoped changes, explicit approval for remote writes, and regression evidence. EffectFence extends that interest into bounded checks of crash, retry, and MCP tool side effects.
AI tooling | Prototype

Data-quality investigation assistant

Built and packaged a VS Code extension with a TypeScript agent layer and Python backend. It brings case review, evidence gathering, and response drafting into one developer workflow.

Grounded investigation
Read-only SQL tools and a local SQLite FTS5 index of database schemas and documentation provide evidence for Copilot-assisted case analysis.
Configurable project knowledge
Added guided setup for queues, permitted databases, and product context, plus reviewer-supplied files and an editable knowledge profile.
Explicit review boundaries
Findings and response drafts require human review. Posting requires approval; credentials use VS Code SecretStorage, and the configurable version uses Windows authentication for SQL access.

Recommendations

From the people I work with.

View on LinkedIn

Education

A foundation in computer science.

Degree

University of Texas at Arlington

Bachelor of Science in Computer Science, May 2024.

Algorithms, operating systems, databases, software testing, AI, machine learning, and networks.

Campus work

From lab support to development

Progressed from Student Lab Assistant to Research Assistant and IT Lab Developer, supporting users before contributing to research and software delivery.

Community

Mobi and ACM at UTA

Contributed as a junior developer with Mobi's software development club and participated in the campus CS community through ACM.

Engineering stack

Tools behind the work.

Used across my professional work and open-source projects.

Build

Languages

C# Python PowerShell SQL TypeScript JavaScript Bash

Ship

Services and data

ASP.NET Core REST APIs SOAP SQL Server PostgreSQL Kafka

Operate

Delivery and reliability

Azure DevOps CI/CD IIS GitHub Actions Docker Fault injection

Augment

Applied AI

GitHub Copilot LLM tool calling RAG LangChain FAISS SQLite FTS5

Recognition

Research and recognition.

Altera Digital Health

Employee Recognition

Recognized for technical ownership, problem-solving, collaboration, and delivery of critical clinical terminology content releases.

VLDB 2025 | Demo paper

MLN-geeWhiz

Co-authored "MLN-geeWhiz: A Dashboard for Supporting Complete Life-Cycle of Complex Data Analysis using Multilayer Networks," presented at VLDB 2025.

Read VLDB demo paper

Academic excellence

Nokia Outstanding Professional CS Student

Received the academic excellence award from the University of Texas at Arlington Computer Science department in both 2023 and 2024.

Times Square, NYC

LinkLeap: URL Shortener

Won 1st prize for the development of LinkLeap, a URL shortener, among more than 300 participants. The project was featured in Times Square in New York City.

HackUTA 2023

Best Use of Streamlit

Won the Streamlit category with the Load-N-Ask team for a document question-answering application.

View award and team

IEEE BigDataService

MLN-Visualizer

Co-authored “MLN-Visualizer: Interactive Visualization of Graphs and Multilayer Network Analysis Results,” presented through IEEE BigDataService.

View IEEE publication

Contact

Let's talk.

I'm interested in backend infrastructure, platform engineering, and developer productivity roles. Let's talk about what your team is building.