• Developed a multi-service hiring simulation platform using NestJS and Prisma, enabling seamless admin-candidate workflows

• Boosted relevance of AI-generated simulation outputs by ∼58% through schema-validated JSON prompts, safe parsing, guardrail design, and fine-tuning on OpenAI models, while deploying multiple AI-powered chatbots

• Improved scalability and reliability of LLM agents by implementing token budgeting, batching, and observability tools, reducing failure cases by 30% and supporting higher user loads

• Collaborated with cross-functional teams to define product requirements and deliver key features

• Designed a fault-tolerant inference framework using TensorRT-LLM’s PyTorch backend, implementing KV cache redundancy across parallel GPU pipelines to ensure high-throughput LLM generation under hardware failures

• Ranked in the top 0.7 percentile globally out of 12620+ teams, developing HFT algorithms within a simulated market environment

• Enhanced forecasting accuracy by applying predictive models like Black-Scholes, Ornstein-Uhlenbeck, and ARIMA

• Consistently generated alpha and optimized PNL using market making, gamma scalping, and ETF arbitrage

• Implemented stop-loss thresholds for risk management, mitigating losses by 22% during high-volatility trading rounds

• Developed a Python app for compliance checks on daily trade data within the Research Infrastructure Group

• Collaborated with the algorithmic trading team, using NumPy, XGBoost, and ARIMA to enhance forecasting accuracy

• Optimized data processing efficiency by 15.2% through the implementation of multi-threading and parallel processing