Blake Burnley

Computer Science @ Georgia Tech

Experience

roles I've had & things I've built

Software Engineer Intern
Jun – Oct 2026 · React · Delphi

Voith Group

Field Report Inspection Application

  • Built a React PWA that digitizes field inspections and auto-generates Word/PDF reports from a corporate template, working fully offline with no backend.
  • Delivered an offline-first inspection reporting tool (React, Vite) featuring in-browser photo annotation and automated Word/PDF export.

Water Press Calculation Application

  • Modernized a legacy nine-variant Excel engineering calculator into a configuration-driven Delphi desktop application, reducing support for a new equipment configuration from a code change to a JSON file.
  • Built an automated Excel-to-JSON extraction pipeline (PowerShell + COM) and a three-tier verification suite (822 compiler-verified assertions plus 844 independent checks) validating every calculation against the source workbook.
  • Diagnosed data-integrity defects in the legacy source that had corrupted headline engineering outputs by 30+ percentage points, and re-architected fragile cell-reference aggregations into a role-tagged classification model.
React Vite PWA Delphi PowerShell Offline-First
Software Engineer Intern
Feb – May 2026 · AWS · Python

Nexus Enablement

AI-Powered Due Diligence Pipeline

  • Architected and deployed an end-to-end AI due diligence automation pipeline processing multi-document investment deals (PDF/DOCX) using AWS Lambda, Claude via AWS Bedrock, and Reducto-based document parsing.
  • Built parallelized extraction workflows with confidence scoring, source excerpt validation, and structured schema generation to automate investment memo creation — including financial analysis, risk assessment, and recommendation outputs.
  • Developed supporting infrastructure including reviewer permissions, audit logging, extraction correction workflows, scheduled cleanup jobs, email notifications, and PDF/DOCX export functionality across four distributed Lambda services.

Cross-Document Intelligence & Citation System

  • Engineered a semantic search and citation verification system enabling reviewers to trace AI-generated outputs back to original PDF/DOCX source documents with inline browser rendering and highlighted citations.
  • Implemented a five-stage text-matching pipeline (exact, sequential, prefix, chunk-based, and fuzzy trigram matching) to robustly locate excerpts across unstructured documents and scanned layouts.
AWS Lambda AWS Bedrock Claude AI Python Reducto Semantic Search

Projects

things I've built

Full-Stack Developer
2025 · Mobile

Community Transportation App

Collaborated with a team of 4 and a non-profit organization (Peoplestown Atlanta) to design a mobile application connecting residents with local community transportation services. Identified community needs through stakeholder engagement, ensuring accessibility, usability, and long-term sustainability. Incorporated user feedback into practical app features to reduce mobility barriers and strengthen local resource access.

Mobile Non-Profit UX Design Community Impact
ML Researcher
2025 · Python · PyTorch

Efficient Financial Forecasting via Knowledge Distillation

Distilled a 700M parameter time-series foundation model (Amazon Chronos-2) into a compact 1.2M parameter Temporal Fusion Transformer for high-frequency intraday stock forecasting. Used a hybrid pinball + MSE loss against teacher soft labels across 20 S&P 500 / Nasdaq-100 symbols. The distilled student achieved 46.89% directional accuracy — outperforming the zero-shot teacher (35%) and persistence baseline (8.32%) — while being ~580× smaller in parameter count. Research conducted for CS 4644: Deep Learning at Georgia Tech.

Python PyTorch Lightning Knowledge Distillation Temporal Fusion Transformer Time Series Chronos-2 Quantitative Finance
Full-Stack Developer
2024 · Python

Dual Meet Lineup Builder

Web-based tool that creates optimal swim meet lineups for single team and strategic dual meet scenarios. Pulls swimmer times using a headless Chrome + Selenium scraper, cleans data with pandas, then runs assignment algorithms to produce the best individual and relay line-ups. Runs in a Docker-containerized Flask API with a responsive front end.

Python Selenium Pandas Flask Docker
ML Engineer
2024 · Python

Sovereign Debt Crisis Prediction System

ML system predicting sovereign debt crises up to 3 years in advance using macroeconomic indicators from 68 countries spanning 1961–2024. Compares Random Forest (0.932 ROC-AUC), Gradient Boosting (0.911 ROC-AUC), and clustering methods. Features engineered using 5-year rolling windows over GDP growth, inflation, credit levels, and reserves.

Python scikit-learn Random Forest Gradient Boosting Feature Engineering
iOS Developer
2024 · Swift

Strain Fitness Tracker

SwiftUI wellness companion pulling workouts, sleep, HRV, stress, and nutrition from HealthKit. Layers recovery dashboards, streak tracking, and a Solo Leveling–inspired Hunter stat sheet on top of on-device CreateML recovery predictions. Core Data stores historical trends alongside guided breathing suggestions and reminders.

SwiftUI HealthKit Core ML Core Data
Full-Stack Developer
2023 · Java

WanderSync

Mobile app enabling real-time collaborative travel planning. Users create shared itineraries, manage destinations, accommodations, and dining plans with secure authentication and live synchronization for group travelers.

Java Firebase Android Agile
Game Developer
2023 · C

Fly Swatter

Fast-paced GameBoy Advance game written in C. The player swats flies within a 15-second timer, competing for a high-score leaderboard. Features name entry, sprite animation, and collision detection built directly on the GBA hardware graphics engine.

C GBA Engine Sprite Animation Collision Detection