DesignDevelopDeploy

Built a Cursor-like AI workstation for traders and quants where autonomous agents collaborated on market research, strategy generation, execution and backtesting, reducing quantitative strategy development time by 40%.

Reduced infrastructure costs by $31.6K+ annually by building a cross-platform x86-to-ARM migration pipeline for C++/Rust services and optimizing AWS compute, ECS, RDS and Aurora workloads.

Built Terraform/Terragrunt modules for multi-cloud trading infrastructure across AWS and GCP using Kubernetes, Helm, ArgoCD and Karpenter.

Deployed and operated MoE LLMs up to 27B parameters using vLLM, tensor parallelism and multi-GPU infrastructure across cloud and customer-hosted environments.

Designed a context optimization framework for long-running AI agents using hierarchical summarization, vectorized memory and Qdrant retrieval, reducing inference costs by 30–60%.

Built evaluation and observability infrastructure using LLM-as-a-Judge pipelines, trace collection and regression detection for production AI systems.

Optimized real-time trading interfaces processing 300+ WebSocket updates/minute using Canvas rendering, bounded queues and memory optimizations, eliminating UI lag and memory leaks.

Built a C++ low-latency pre-trade risk engine supporting $100M+ trading volume, enforcing margin checks, leverage limits, exposure controls and duplicate order prevention.

Built real-time post-trade analytics and reconciliation infrastructure validating executions, balances, positions and fees while detecting exchange-to-internal state mismatches.

Provided primary operational support for cloud and bare-metal trading infrastructure, performing incident triage, root cause analysis and restoring production trading systems during critical outages.

Implemented algorithmic execution and risk management workflows including TWAP, VWAP and trailing stop-loss strategies.

Developed exchange connectivity and execution infrastructure integrating REST, FIX and WebSocket APIs for market data ingestion and order routing across 20+ trading venues.

Reverse engineered undocumented exchange workflows through API analysis, source inspection, protocol specifications and blockchain interactions, recreating margin and leverage operations where documentation was incomplete.

Optimized analytics workloads through PostgreSQL query tuning, indexing strategies, rolling aggregations and Redis caching.

Designed a secure credential brokerage platform using Go and OpenBao, providing audited exchange credential management, account allocation and access control.

Reduced frontend memory consumption by 60%, build times from 25 minutes to 3 minutes and release bugs from 8 to 0 per cycle by redesigning frontend architecture and introducing 600+ automated tests.

Built CI/CD, end-to-end testing and stress-testing frameworks using GitHub Actions, Playwright, remote caching and high-frequency workload simulation.