Quantitative researcher and AI systems engineer focused on cross-asset forecasting, financial time series, and machine-learning infrastructure. Built a C++ event-driven prediction engine with reported 36.5× cumulative growth and a 2.96 Sharpe ratio over an 11-year backtest. Developed Python/PyTorch/TensorFlow forecasting workflows, real-time market-data pipelines, and multi-agent research systems. Brings strong foundations in statistical experimentation, algorithms, data structures, and financial research software.
- Core Strengths: Multi-agent AI systems architecture, cognitive architectures for LLMs, quantitative forecasting and predictive modeling, institutional-grade trading frameworks.
- Quantitative Finance: Quantitative research, financial time-series analysis, cross-asset forecasting, portfolio allocation, Monte Carlo methods, market microstructure, strategy backtesting, execution modeling, OHLCV ingestion, return-sign synchronization, collective entropy, RMT covariance denoising, Barra risk attribution, CVXPY optimization, Almgren–Chriss execution.
- Languages: Python, JavaScript, SQL, Go, C++; algorithms, data structures, statistical experimentation, computational modeling.
- AI/ML: PyTorch, TensorFlow, LangChain, neural networks, forecasting models, LLMs, cognitive architectures, multi-agent systems, machine learning, Monte Carlo methods, local inference.
- Frameworks & APIs: REST APIs, Groq API, Ollama, FastAPI, LangChain, NetworkX, ChromaDB, SQLite, authentication systems, PostgreSQL, CUDA.
- Systems: Multi-agent AI systems, distributed reasoning, workflow automation, memory state machines, two-stage reranked retrieval (RAG), event-driven architectures, secure systems, algorithms, data structures.
- Data Engineering: Web scraping, real-time news aggregation, data parsing, OHLCV ingestion, multi-asset alignment, timezone normalization, data preprocessing.
- Domains & Infrastructure: Predictive modeling, AI systems, quantitative research, financial time-series analysis, cross-asset forecasting, portfolio allocation, financial industry experience, infrastructure technologies, early-stage ventures, start-up leadership, AI inferencing.
- Hardware: SystemVerilog, FPGA-oriented design, edge-AI SoC architecture.
- Graphics: Computer graphics.
Leading technology strategy and engineering at a stealth startup, building core systems and driving product development across AI, software, and emerging technologies.
Worked on computer graphics and graphics systems, contributing to graphics infrastructure and rendering projects.
Founded a cybersecurity startup focused on application and website security, building automated security analysis and vulnerability detection systems.
Conducted independent research across quantitative finance and AI/ML, building predictive models, trading systems, intelligent architectures, and research infrastructure. Developed systems including Ouroboros, Athena, and Oron.
Pairs causal temporal convolutions with a dynamic correlation graph for cross-sectional ranking; the portfolio goes long the top decile and short the bottom decile with a one-day execution lag and transaction costs. Walk-forward development Rank IC / Net Sharpe: 10D 0.0923 / 0.556; 20D 0.1159 / 0.521; 60D 0.0807 / 0.156. The post-2023 holdout is not included.
Uses cross-sectional entropy and return-sign synchronization to derive a defensive portfolio-allocation signal.
Studies how disclosure, crowding, arbitrage, execution costs, and regime change affect implementable alpha persistence.
Combines time- and frequency-domain prediction with regime conditioning, probability calibration, and selective cross-asset evaluation.
- Ito — Trading-system foundation with deterministic simulation, risk controls, venue connectivity, FPGA-oriented gating, and production-readiness validation.
- HPT-DTE-CAFP — Harmonic phase-transformer and decision-tree ensemble for cross-asset financial prediction.
- Zerene — Discrete-event exchange simulator with FIFO matching, fixed-point ticks, multi-hop latency, OFI-conditioned Hawkes flow, and incremental risk checks. Repository benchmark: 33,000–67,000 ops/sec and 12.2–16.5 μs median fills on a stated 100,000-order single-threaded CPython 3.14 workload.
- ORX-1 — SystemVerilog edge-AI SoC reference design with RV32I control, sensor pipeline, NPU integration, secure boot, and formal interface checks.
- RAMP — Quant research/execution engine using causal regimes, RMT covariance denoising, Barra risk attribution, CVXPY optimization, and Almgren–Chriss execution.
- CESP — Cross-asset risk allocation using collective entropy and return-sign synchronization.
- Klint-32M — 32M-parameter, 10-layer causal Transformer for OHLCV generation; three-stream residual vector quantization (price path, range, activity), 256-bar context, and geometry-constrained decoding. Repository reports v2 fine-tuning across 101 asset regimes with PnL-weighted cross-entropy and directional penalties, reducing validation loss from 8.80 to 2.81. This is a model-development metric, not a profitability result.
- Mosaic: High-frequency L2/L3 order-book matching, multi-level order-flow imbalance (OFI), FIFO queue simulation, and purged walk-forward machine learning.
- Sizon: Evolutionary alpha research using NSGA-II expression search, chronological purge/embargo validation, and microstructure simulation.
- Published serialized digital fiction on WEBTOON, managing long-form narrative development and episodic publishing workflows.
- Explored systems-oriented storytelling and speculative-technology worldbuilding through serialized web publication.
- Stanford Education Everywhere (SEE) participant.
- Independent coursework: Harvard CS50P and CS50AI; MIT OpenCourseWare in AI and finance; WorldQuant University programs focused on quantitative finance and data-driven analysis.