Backend systems · Automation · Applied AI
I build reliable software for complex, real-time operations.
Software engineer with 4+ years of experience designing distributed platforms, high-throughput backend services, and production automation. My work spans broadcast operations at a Sony Group company, multi-customer SAT automation, and independent AI/ML systems.
Production throughput
360–380+
document requests per hour
Distributed footprint
100+ sites
real-time operational monitoring
About
Engineering depth across systems, automation, and AI.
I specialize in turning operationally difficult workflows into maintainable software: systems that move data reliably, recover from interruption, expose clear APIs, and remain observable in production.
At Crispin Corporation, a Sony Group company, I helped build and evolve distributed monitoring, synchronization, media-asset, and real-time broadcasting services used across more than 100 sites. My current freelance work focuses on Python and Playwright automation for Mexico's SAT platform, including encrypted credential handling, parallel browser execution, recovery workflows, and Linux production operations.
Alongside backend engineering, I am developing deeper expertise in applied AI through independent projects in computer vision, graph neural networks, retrieval-augmented generation, and MLOps.
Experience
Production systems built for real operational constraints.
Freelance Software Developer
Valarix AI
- Built a production Python and Playwright platform for SAT CFDI retrieval and e.firma/FIEL validation, integrating backend REST APIs with encrypted credential workflows.
- Automated Opinión de Cumplimiento and Constancia de Situación Fiscal retrieval with FTP delivery, monthly scheduling, state tracking, and interrupted-run recovery.
- Implemented parallel multi-customer processing with
ProcessPoolExecutor, preserving sequential document flow within each customer session. - Deployed and operate the system on a DigitalOcean Linux droplet through PM2, structured logging, rotation, and live monitoring, sustaining approximately 360–380+ document requests per hour.
Software Developer
Crispin Corporation · Sony Group Company
- Designed an end-to-end monitoring platform spanning CoreAgent data collection, RabbitMQ publishing, multi-agent backend aggregation, configuration, filtering, and operator-facing visualization across 100+ sites.
- Improved high-throughput pipelines serving 10–15 channels per site and approximately 5,000+ events per channel through concurrency, caching, batching, and dashboard-specific filtering.
- Engineered hub synchronization and real-time JIP broadcasting services that maintained consistency while supporting last-minute changes during live broadcast operations.
- Launched the Assetbase Service and led config-service architectural improvements that reduced render latency and streamlined media-asset operations.
Software Developer Intern
Incluvie
Delivered responsive product features in React, Hooks, and TypeScript while optimizing performance and integrating changes safely into an existing codebase.
Graduate Teaching Assistant
UNC Charlotte
Supported Algorithms and Data Structures under Dr. Qiong Cheng, including instruction, technical guidance, debugging support, assessment design, and evaluation.
Selected work
Systems designed around scale, latency, reliability, and explainability.
Production work is identified separately from independent AI/ML projects.
Production platform · Crispin
Distributed Real-Time Monitoring Platform
Architected a multi-layer platform for live operational visibility: CoreAgent collection, RabbitMQ event transport, backend aggregation and filtering, configuration management, and operator-facing dashboards.
- 100+ sitesdistributed operational coverage
- 5,000+ eventsper channel at high volume
Node.jsRabbitMQGraphQLC#Knockout.js
Production platform · Valarix AI
SAT Document Automation & FIEL Validation
Built a fault-aware automation service for authenticated SAT workflows, combining secure credential handling, concurrent browser sessions, API coordination, FTP delivery, recurring schedules, and resumable state.
- 360–380+document requests per hour
- Recovery-firstinterrupted-run continuation
PythonPlaywrightREST APIsPM2DigitalOcean
Production services · Crispin
Hub Synchronization & Live Broadcast Services
Developed synchronization services that preserved consistency across distributed hubs and real-time JIP workflows designed for low-latency, last-minute broadcast changes without disrupting live operations.
Node.jsExpressGraphQLRabbitMQC#
Production service · Crispin
Assetbase Media Service
Led the design and launch of a media-asset service that improved organization, centralized asset operations, and reduced friction across content-management workflows.
Node.jsPythonFlaskREST
Independent AI/ML projects
Research-oriented systems with measurable technical outcomes.
Computer vision
Edge-Optimized Multi-Object Tracking
Combined YOLOv10 and ByteTrack for persistent multi-object identity across live video, then explored TensorRT INT8 quantization for edge deployment.
15 → 52 FPS3.4× inference speedup with minimal accuracy loss
PythonYOLOv10ByteTrackTensorRTFastAPIDocker
Graph machine learning
Relational GNN Fraud Detection
Modeled 100K+ financial entities and relationships as a heterogeneous graph, implemented an R-GCN for imbalanced fraud classification, and used GNNExplainer for structural interpretability.
+14% F1minority-class improvement over the comparison baseline
PyTorch GeometricNetworkXGNNExplainerScikit-Learn
Retrieval systems
Asynchronous RAG with Hybrid Retrieval
Designed a FastAPI ingestion and retrieval framework with sparse-dense search, semantic Redis caching, JWT authentication, and token-bucket rate limiting.
92% precisionhybrid retrieval performance in project evaluation
FastAPIRedisVector embeddingsJWTAsync Python
MLOps
Automated Retraining & Drift Pipeline
Created a demand-forecasting workflow for data ingestion, feature engineering, experiment tracking, versioned datasets, drift detection, and CI/CD-triggered retraining.
50+ runstracked experiments and hyperparameter trials
AirflowDVCMLflowEvidentlyAIGitHub Actions
Earlier academic and full-stack work
LendingTree Site Visitor Behavior Visualization
Developed an interactive analytics dashboard for exploring LendingTree visitor behavior across U.S. states and lending-product categories. Visualized traffic volume, acquisition channels, offer engagement, signup activity, popular user actions, and form completion patterns to support geographic and product-level behavioral analysis.
D3.js
JavaScript
Data Visualization
Behavioral Analytics
View live visualization
Secure Digital Document Locker
Designed a permissioned document-sharing platform for students and universities as a selected participant in National Smart India Hackathon 2017.
Access controlDocument workflows
Django Personal Publishing Platform
Implemented authentication, profile management, and content publishing in a complete Django web application.
View repositoryAnime World
Built an MVC application for discovery, ratings, watch-state tracking, favorites, and genre organization.
View repository
Technical toolkit
Broad enough to ship end to end; deep where reliability matters.
Languages
Python, JavaScript, C#, TypeScript
Backend & APIs
Node.js, Express.js, FastAPI, Flask, Django, GraphQL, REST APIs
Distributed Systems
RabbitMQ, event-driven services, concurrency, synchronization, caching, batch processing
AI & Machine Learning
PyTorch, PyTorch Geometric, Scikit-Learn, YOLOv10, TensorRT, vector retrieval
Data & Infrastructure
SQLite, MySQL, Redis, Docker, DigitalOcean, PM2, GitHub Actions
MLOps & Observability
Airflow, DVC, MLflow, EvidentlyAI, structured logging, production monitoring
Frontend
React, Hooks, Knockout.js, HTML5, CSS3, Axios, jQuery
Engineering Practices
Object-oriented design, SDLC, Agile delivery, API design, performance diagnosis
Education & highlights
Academic foundation reinforced by competitive problem solving.
Aug 2018 – May 2020
Master of Science in Computer Science
University of North Carolina at Charlotte
Graduate Teaching Assistant for Algorithms and Data Structures. Relevant study included algorithms, visual analytics, intelligent systems, and network-based application development.
Queen City Data Science Hackathon
Red Ventures · 2019
Data-science solution addressing a substance-abuse problem.
Recognition
2nd Place — Queen City Data Science Hackathon
Our team selected and analyzed a Charlotte-area dataset focused on substance abuse and homelessness, then trained a machine-learning model to identify meaningful patterns and support data-informed decision-making.
We also developed an interactive engineering layer that allowed users to query the model directly and enabled downstream applications to consume its predictions. The project was completed with mentorship from experienced data scientists and engineers in the Charlotte technology community and earned second place at the 2019 Queen City Data Science Hackathon.
Machine Learning
Data Analysis
Model Integration
Decision-Support Systems
Reinvent the Wheel Hackathon
American Tire Distributors · 2019
Applied machine learning to a tire-industry forecasting use case.Queen City Data Science Hackathon
Red Ventures · 2020
Developed a data-science solution focused on the opioid crisis.National Smart India Hackathon
Selected participant · 2017
Designed a secure, permissioned digital document locker.
Contact
Let’s discuss engineering problems worth solving.
For software engineering, backend, distributed-systems, automation, or AI-focused opportunities, reach me directly or use the form.