Sparsh Verma

• Founded and built ZitBoard, an AI-powered business operating system that
unifies sales, hiring, and operational data into a single source of truth, enabling
faster and more informed business decisions.
• Architected and shipped a production-grade software platform solo from zero
to 100+ active users; applied rigorous software design principles, modular ar
chitecture, and system-level performance optimization across the full stack.
• Built and optimized AI inference pipelines using TensorRT-aligned deployment
practices; focused on minimizing latency, maximizing throughput, and ensuring
model reliability under real-world load.
• Designed automated testing and CI/CD frameworks to enforce code quality,
system correctness, and continuous performance monitoring across all produc
tion components.

• Designed, trained, and deployed ML models in Python/C++ to process and
classify 20, 000+ daily transactions across 127 business units; applied deep
learning techniques and feature engineering to maximize algorithmic accuracy
and throughput.
• Engineered PySpark distributed data pipelines processing ~4 TB/month; lever
aged parallel computing techniques for high-throughput, low-latency data pro
cessing at enterprise scale.
• Built anomaly detection algorithms across millions of records using advanced
statistical and ML methods; performed in-depth analysis and root cause iden
tification to surface ~$10, 000+ in quarterly loss avoidance.
• Communicated complex algorithmic findings and model performance analysis
directly to business stakeholders, translating technical depth into actionable
decisions.

• Optimised reinforcement learning models by ~20% through systematic archi
tecture tuning, hyperparameter analysis, and performance profiling of ML al
gorithms on cloud infrastructure.
• Contributed to software design reviews, code refactoring, and algorithmic per
formance optimisation across production ML systems.