Software Engineering | Applied AI | Distributed Systems | Open Source
Three years of engineering across production software and applied AI. At Volkswagen I learned that most of the job is not making something work once, it is making it keep working. The same properties that mattered in dealer platforms matter in ML systems: correctness, observability, and behavior under conditions you did not anticipate. Which is why I find a model's failure modes more interesting than its benchmark scores. MS in Artificial Intelligence at WPI.
Fine-tuning, evaluation, and agentic workflows. My PayPal-sponsored graduate capstone fine-tuned Gemma-3 4B with LoRA on a 485k-row credit dataset, reaching ROC-AUC 0.829 and beating XGBoost and the larger Qwen3-8B. The more useful finding was negative: the fine-tuned LLMs collapsed out-of-domain while tabular foundation models held. Placed 1st.
At Volkswagen: cut billing query latency from 30s to 10s, reduced production runtime errors by 50 percent through REST API redesign and automated CI/CD, and resolved 150+ critical issues across the stack.
The area I am deliberately growing into. So far this has meant microservices and multi-region localization at Volkswagen, containerized workloads on AWS EC2, Lambda, and S3, HPC/SLURM for training jobs, and Prometheus and Grafana observability across 10+ automation workflows. Currently working through consensus, replication, and failure modes in more depth.
Everything I build outside of work lives here, public and documented.
Jerai Bug-fixing agent. Reads an issue, runs tests, writes a valid patch in 10 to 15 seconds. RAG over ChromaDB for similar bug patterns.
Llama 3.3 Cerebras Docker MCP Flask
CampusCashControl Finance audit platform with a production ingestion pipeline, RBAC, and JWT auth.
Django Vue PostgreSQL AWS
WPI IT Automation — LLM workflows handling retrieval, validation, and multi-step orchestration across 50+ support tickets.
LangChain LangGraph Docker AWS
Languages Python, Java, C#, Go, TypeScript, SQL, NoSQL
AI/ML PyTorch, LangChain, LangGraph, CrewAI, Microsoft Autogen, LoRA/PEFT, RAG, Model Context Protocol, XGBoost, Scikit-learn, Pandas, NumPy
Backend Django, FastAPI, ASP.NET Core, Spring Boot, microservices
Data PostgreSQL, MongoDB, pgvector, ChromaDB, Redis
Infrastructure AWS, Azure, Docker, Kubernetes, CI/CD, Linux, HPC/SLURM, Prometheus, Grafana, LangSmith, LangFuse