Vaigarai Sathi

Computer Science Graduate Student

Passionate about quantitative research, machine learning, and building scalable systems. Currently pursuing MS in Computer Science at NYU with expertise in trading algorithms, ML infrastructure, and full-stack development.

View Projects Get In Touch
Vaigarai Sathi

MS Computer Science

GPA: 3.96/4.0

About Me

I'm a dedicated Computer Science graduate student at New York University with a strong foundation in quantitative research, machine learning, and software engineering. My journey spans from developing advanced trading algorithms to building scalable ML infrastructure.

I'm passionate about solving complex problems at the intersection of finance and technology, with expertise in real-time systems, cloud architecture, and machine learning model deployment. My research interests include graph learning and reinforcement learning.

Languages

Python Java JavaScript C++ C HTML5 CSS3

Frameworks & Tools

PyTorch TensorFlow React Spring Boot Node.js Flask

Cloud & DevOps

AWS Google Cloud Docker Kubernetes Jenkins Git

Experience

Quantitative Development Intern

Blockhouse Labs, Inc.

June 2025 - Aug 2025

Engineered advanced trading strategies and algorithms for a multi-venue Smart Order Router (SOR), optimizing execution across equities, crypto, and derivatives.

  • Implemented affine alpha integration with TWAP-based scheduler using convex optimization
  • Designed real-time policy engine with <10ms response time for market microstructure signals
  • Reduced adverse selection costs by 23% during high-volatility periods
  • Accelerated SOR decision-making by 17% with ML-based lookup model

Machine Learning Intern

Onward Assist

Jan 2024 - July 2024

Led cloud infrastructure migration and ML workload optimization for pathology applications.

  • Engineered scalable container orchestration on AWS ECS for ML workloads
  • Led migration from AWS to GCP with Kubernetes Engine architecture
  • Enhanced pathology application with DICOM compatibility for medical imaging
  • Developed comprehensive format conversion pipeline using Cloud Healthcare API

SDE Intern

Effigo Global

Jan 2023 - July 2023

Modernized legacy applications and improved development workflows through automation.

  • Modernized legacy Spring Boot applications and developed RESTful APIs
  • Managed AWS instances and Docker container deployments
  • Architected reusable mailing service with open-source editors
  • Optimized Jenkins CI/CD pipelines with automated builds and notifications

Featured Projects

Spatio-Temporal Graph Learning for Covariance Forecasting

PyTorch, Transformers, Graph Neural Networks

Advanced financial modeling using graph-based representations of S&P 500 markets with spatio-temporal transformers for high-resolution covariance forecasting.

  • Constructed dynamic graph representations using S&P 500 assets as nodes with multi-source features
  • Developed Spatio-Temporal Graph Transformer combining GATv2 and Transformer encoders
  • Enforced positive semi-definiteness via low-rank factorization and spectral regularization
  • Optimized log-likelihood-based loss with Ledoit-Wolf-style shrinkage for robust risk modeling

Adaptive GPU Hash Table

C++, CUDA, GPU Programming

Engineered a high-performance, header-only hash table in CUDA using Cooperative Groups to parallelize probing sequences and Structure of Arrays to optimize memory coalescing.

  • Implemented linear probing with tombstone-based deletion and an adaptive resizing/compaction mechanism
  • Achieves a 1.8x speedup in read performance after compaction while significantly reducing GPU memory footprint
  • Optimized memory access patterns for high-throughput GPU operations
  • Designed for seamless integration into CUDA applications with minimal dependencies

Cooperation Dynamics in Multi-Agent Systems

Unity, C#, PyTorch, Game Theory

Research on promoting cooperation in multi-agent systems using game-theoretic approaches, published at IROS 2023 and SAE International Journal.

  • Presented research paper at MAD-Games Workshop at IROS 2023
  • Published journal article in SAE International Journal of Connected and Automated Vehicles
  • Leveraged mean-field game theory for equilibrium solutions in infinitely large agent sets
  • Validated practical capabilities using various simulation environments

Predicting Probability of Default for Banca Massica

Python, Scikit-learn, XGBoost

Built a 12-month Probability of Default (PD) model for business loans using firm-level financial statements, enabling risk-based pricing and underwriting decisions.

  • Engineered financially grounded features across profitability, leverage, liquidity, activity, size, and industry
  • Built a champion-challenger framework (Logistic Regression → XGBoost) with monotonicity constraints
  • Delivered explainable PD estimates suitable for credit approval workflows, stress testing, and borrower-level risk assessment

Let's Connect

I'm always interested in discussing new opportunities, research collaborations, or just having a chat about technology and innovation.

Email LinkedIn GitHub