Nihal Tahariya

Designed and implemented a custom lock-free ring buffer, achieving a 6X latency improvement over Boost SPSC, minimizing packet loss in real-time data streams. Enhanced cancel rejection logic to include broker-specific reasons, improving system accuracy and client trust. Redesigned memory allocation in latency-critical modules, eliminating heap contention and boosting throughput during peak market load. Implemented a publishersubscriber model within the event-driven architecture to handle real-time market data dissemination.

Optimized equities exchange connectivity adapter with TCP-level improvements, cutting latency by 10% and improving market data reliability. Built an L2 order book to manage high-frequency pending orders, ensuring accuracy in volatile trading conditions. Provided production support, diagnosing and fixing low-level bugs in high-load environments.

Built an end-to-end image processing pipeline in C++, powering analytics such as zone-wise time and penalty tracking. Extended the camera calibration system to support distributed multi-sport environments, ensuring accuracy across varied setups. Developed a GUI-based calibration tool to streamline workflows for field operators.

Engineered a snapshot recovery mechanism for market data adapters to guarantee data integrity from UDP multicast feeds. Designed and implemented an ultra-low-latency limit order book, processing orders in under 300ns using cache-friendly data structures. Established exchange connectivity and built an OMS, enabling real-time order execution with high resilience. Applied design patterns to enhance scalability and maintainability of the trading system. Adopted Test-Driven Development (TDD) in Python, achieving high unit test coverage for mission-critical components.

Built and deployed a computer vision system for real-time defect detection on 1500 packets/min cigarette lines using Jetson Nano. Deployed deep learning inference pipelines on embedded systems, cutting defect escape rates significantly. Optimized acquisition-to-inference pipeline for low-latency streaming anomaly detection.