Career
Experience
Systems integration, backend, data, and ML work across CU Boulder, Purdue, and L3Harris.
Systems Integration Engineer
University of Colorado Boulder · Institute of Behavioral Science
- Building a FastAPI, React/TypeScript, and PostgreSQL support-operations platform around a current 10,803-ticket Freshdesk-backed corpus, with support-group IDs recorded for 96.7% of records
- Developed and evaluated an unreleased BERT routing candidate on 6,137 training tickets, reaching 82.63% sealed-test accuracy and 0.7563 macro-F1 while keeping deployment separate from evaluation
- Built a React, Vite, and Recharts dashboard that separates source intake from reportable volume and exposes busiest weekdays and peak hours independently
- Reduced the analytics visualization payload by 61% raw and 66% gzip by replacing Plotly with Recharts and accessible chart tables
- Built a PowerShell tool that scans the IBS organizational unit in Active Directory and removes group memberships, reducing a typical offboarding cleanup from about 15 minutes to under one minute
- Managed Windows Autopilot and Jamf Pro enrollment for 50+ endpoints under Secure Compute requirements
January 15, 2025 – May 2026
Research Assistant — Indian Parliamentary Data
Princeton University
Functional title · Paid Princeton appointment · End-to-end parliamentary-data engineering
- Paid directly by Princeton under a separate appointment to engineer the end-to-end workflow after the Studio Lab evaluation phase
- Built the scraping logic for dynamic parliamentary archives, including pagination handling and resumable SQLite state
- Orchestrated collection, speech and translation processing, PDF text extraction, normalization, and structured outputs for downstream analysis
- Processed 40+ GB of campaign audio and thousands of parliamentary documents while retaining reviewable intermediate artifacts and restart points
Lead Technical Research Assistant
University of Colorado Boulder · Institute of Behavioral Science
Separate concurrent IBS appointment · Independent research workstreams
- Worked with PhD researchers to build a PyTorch and ResNet50 pipeline for studying visual bias in New York Times COVID-19 imagery. Added OpenCV and scikit-image preprocessing with z-score normalization and channel-wise intensity scaling
- Used a truncated ResNet50 to extract image embeddings, then grouped them with k-means for analysis without labeled data
- Built Selenium and BeautifulSoup pipelines for 15+ years of legislative data, with JSON checkpoints for long-running jobs
- Built a Python wrapper for the LegiScan API with caching and rate-limit handling, cutting redundant requests by 40% during bulk ingestion
Undergraduate Research Assistant, Satellite Telemetry Data Analysis
The Data Mine of the Rockies · Purdue University and L3Harris
Purdue research program · L3Harris industry collaboration
- Worked directly with L3Harris senior engineers through Purdue’s Data Mine of the Rockies on satellite telemetry cybersecurity research
- Co-authored and presented “L3Harris: Satellite Cyberattack Detection” at the inaugural Data Mine of the Rockies Symposium in Colorado Springs to engineers and stakeholders from the U.S. Space Force, L3Harris, Lockheed Martin, Northrop Grumman, CrowdStrike, and other aerospace and cybersecurity organizations
- Trained Isolation Forest and LSTM models on synthetic NOS3 satellite telemetry to detect and classify cyberattack anomalies
- Used Wireshark to identify affected devices and indicators of compromise in simulated satellite communications
- Mapped space-cyber scenarios to the SPARTA Matrix and MITRE ATT&CK frameworks, then generated synthetic telemetry for multiple attack types
January 21 – April 13, 2024
Research Assistant, Data Engineering and ML
Studio Lab · University of Colorado Boulder
Paid CU Boulder appointment · Speech and translation tool evaluation
- Designed comparative speech-to-text evaluations using word error rate to study Deepgram, Google Speech-to-Text, Amazon Transcribe, and Azure Speech for the research workflow
- Set up parallel translation comparisons across Google Cloud Translation, Amazon Translate, and Azure Translator
- Researched service configuration and language-adaptation options for political names and domain-specific vocabulary
- Documented tool behavior and tradeoffs so the project could choose components based on workflow fit rather than vendor claims
- Began the Indian parliamentary-data research under this paid Studio Lab appointment; the later Princeton appointment covered the end-to-end data-engineering phase
