Yale University
Algorithms Research Intern
Designed and implemented a novel parallel graph algorithm in C++ that updates clique counts after batches of edge insertions and deletions, avoiding a full recount after each batch.
John Quevedo
Software Engineer
I study Computer Science and Mathematics at Yale University, focusing on backend engineering, distributed systems, and machine learning.
Currently
Featured projects
CI Failure Capture and Replay
Reproduces CI failures by preserving both code and runtime state. Built with idempotent jobs, lease-based queues, checkpointed workers, and isolated replay containers.
Completed 172 of 172 controlled replays.
- Go
- Python
- PostgreSQL
- Docker
- Kubernetes
- GitHub Actions
Online Payment Repair
Traces corrupted payments to responsible code or configuration versions, then generates auditable repair plans with concurrency protection.
Repaired 100,000 records at 865 records per second with no ledger imbalances.
- Java
- Kafka
- PostgreSQL
- Spark
- Kubernetes
CPU and GPU Inference Tuning
Selects hardware-specific inference configurations by testing quantization, pruning, compilation, batching, and runtime choices under quality and latency limits.
Reduced median latency by 79.6% with no measured quality loss.
- Python
- PyTorch
- CUDA
- TensorRT
- ONNX Runtime
- Ray
Training Data Auditor
Combines six label, similarity, and model-based signals into explainable review queues and auditable dataset patches.
Profiled 127,600 rows in 9.8 seconds and achieved 91.3% adjudicated precision.
- Python
- PyTorch
- Hugging Face
- Spark
- FAISS