SYSTEMS × AI × GPU COMPUTING
M.S. Computer Science & Engineering @ University of Michigan
I build and study systems that make AI and large-scale computation faster, more reliable, and easier to operate. Previously, I spent four years at Goldman Sachs building production systems for financial data and regulatory reporting.
My background spans production software engineering, large-scale data systems, GPU computing, reinforcement learning, and LLM infrastructure. At Goldman Sachs, I worked on cloud-native data pipelines and regulatory systems operating at financial-institution scale.
At Michigan, I am extending that systems foundation toward AI: from dependency-aware execution for LLM-driven knowledge graphs, to multi-agent reinforcement learning, to high-performance CUDA implementations. I enjoy problems where the algorithm matters, but the system that executes it matters just as much.
LLM infrastructure, execution, agentic workflows
CUDA, GPU architecture, HPC, optimization
MARL, MAPPO, PPO, vision, VLMs
Distributed processing, cloud data, DAGs
Engineered a dependency-aware execution framework that maps semantic dependencies into a DAG to isolate conflicts and enable safe parallel batching for LLM-driven knowledge graph updates.
Developed a multi-class diagnostic system using EfficientNetB2 and Vision Transformer models, with differential learning rates and a custom PyTorch wrapper.
Engineered a client-side personalization engine that learns user gain preferences in real time, with local model persistence and genre-based initialization.
Led architecture of a Spark, AWS Glue, and Snowflake-based insurance calculator processing 6M+ deposit accounts daily under a 20-minute SLA. Architected one of the firm's first cloud-native AWS data pipelines for US deposit information, including customer PII, and supported FDIC Part 370 compliance.
Automated and streamlined German Deposit Protection and European regulatory reporting systems using Java, SQL, REST APIs, and Sybase IQ. Improved the EdB regulatory rating from E (Fail) to B and reduced quarterly reconciliation effort by 60%; received the GS Controllers India Award.
Developed a full-stack reconciliation platform using React and Java Spring, automating reconciliations between General Ledger and Regulatory Reporting data for 500+ Controllers and saving 800+ person-hours each quarter.
M.S. Computer Science & Engineering
GPA: 3.83 / 4.00
Applied GPU Programming · Advanced Scalable Systems for Agentic AI · Reinforcement Learning · Advanced Computer Vision · Probability & Random Processes · Computational Complexity (A+) · Human-AI Interaction (A+) · Learning Enabled Systems
B.E. Computer Science
GPA: 3.98 / 4.00 (9.77 / 10.00)
Data Structures & Algorithms · Operating Systems · Parallel & Distributed Computing · GPU Computing · Cloud Computing · Artificial Intelligence · Image Processing
LET'S BUILD SOMETHING DIFFICULT