MS Graduate @ Arizona State University  |  Software & ML Engineer  |  Full-Stack Builder

I recently graduated with my MS in Data Science, Analytics and Engineering from Arizona State University (GPA: 3.80), with hands-on experience building production-grade software systems, ML pipelines, and full-stack applications.

My work sits at the intersection of software engineering, applied ML, and data engineering. I care about building things that actually work in production, not just in notebooks.

• 🏢 Former Software Development Engineer Intern @ Sentari AI (New York)
• 🔧 Built a self-healing SRE Agent from scratch that resolves CI/CD failures autonomously with a 95% success rate
• 🛠️ Shipped real products across backend, frontend, and ML across multiple internships and projects
• 📍 Based in Tempe, AZ — open to relocation

• Architected REST microservices in Python integrating AI pipelines, processing 30+ simultaneous data streams and reducing inference latency by 21%
• Launched a full-stack React + FastAPI monitoring dashboard enabling real-time anomaly detection, cutting incident response time by 35%
• Engineered responsive frontend features in TypeScript and React, contributing to 200% month-over-month platform growth
• Built an offline processing pipeline that safely queued data and delivered it reliably once connectivity was restored
• Secured a full-stack login page with OAuth2 authentication in TypeScript & React, improving token verification speed by 27%

• Built scalable backend systems in Python, Django, and PostgreSQL, accelerating system response times by 18%
• Streamlined CI/CD pipelines using Docker and GitHub Actions, cutting manual release errors by 30%
• Traced and resolved complex REST API and microservices integration issues using pytest, reducing systemic defects by 25%
• Formulated API specifications and architecture diagrams supporting scalability of 5+ core platform features

• Reduced data retrieval latency by 28% through refactoring core database access layers and applying aggressive caching strategies
• Deployed a real-time analytics monitoring architecture aggregating telemetry data across 1,000+ concurrent user sessions

LangGraph · Groq Llama 3.3-70B · FastAPI · Streamlit · PyGithub · LangSmith · Python AST

Multi-agent AI system for automated incident response. Built from scratch on nights and weekends because the problem was worth solving. V1 is complete and the core self-healing loop works end to end.

• Multi-agent orchestration via LangGraph — Investigator and Mechanic agents with explicit handoff and communication tracking
• Self-correcting loop — up to 3 attempts with validation feedback, mimicking how human engineers debug
• 95% success rate after self-correction, resolving incidents in 30-60 seconds at $0.02-0.06 per fix
• Full observability via LangSmith decision tracing and GitHub Actions logs
• Safe by default — iteration limits, AST-based code validation, human approval required before merge

FastAPI · Redis · React Native · MongoDB · RAG · LLM Evaluation

End-to-end RAG-powered mental health platform using transformer models, embedding models, and vector databases. Processes thousands of daily queries with 91% classification accuracy and sub-500ms latency, with a built-in LLM evaluation pipeline for continuous quality improvement.

PyTorch · XGBoost · Flask · React · MLflow · SHAP · Dask

MS capstone project. End-to-end predictive maintenance system estimating Remaining Useful Life across turbofan engines, batteries, and electrolytic capacitors, combining all three into a unified Go/No-Go fleet readiness decision.

• Engine Bi-LSTM achieved RMSE of 12.77 cycles on the NASA C-MAPSS benchmark
• Battery XGBoost reached R² of 0.882 on a 50-cycle horizon
• Capacitor model predicted end-of-life within a 6 to 9 day scheduling window
• Deployed through a Flask REST API and React dashboard with per-aircraft RUL trends and interactive what-if predictor

PyTorch · U-Net · Sentinel-2 · Feature Engineering · Deep Learning

High-performance deep learning segmentation pipeline on satellite imagery. Applied weighted loss functions and feature engineering to raise minority class recall by 14%, achieving 95.95% accuracy and 92.79% mean IoU.

Python · Cosine Similarity · Embedding-based Retrieval · Ranking

Candidate retrieval and ranking pipeline using embedding-based collaborative filtering. Improved precision by 15%, reduced runtime complexity by 22%, and boosted engagement by 12%.

Open to Software Engineer, ML Engineer, and Data Engineer roles — available immediately