Driving Search & Discovery Through Data.
Experimentation, Statistical Modeling, and AI-Powered Analytics for Marketplace Optimization.
Professional Portfolio
Marketing Experimentation Platform
AI-powered chatbot for marketing experiment analysis. Implements binomial tests, Z-tests, chi-square, and post-hoc power analysis. Automated statistical winner determination with confidence intervals.
- Experiment Analysis: Binomial, Z-test, Chi-square tests
- Power Analysis: Post-hoc statistical validation
- Automated Decisioning: AI-assisted winner determination
AI-Powered Analytics Platform
Multi-workspace BI platform with natural language queries via Databricks Genie API. Enables semantic exploration of complex datasets with contact/account lookup, buyer journey analysis, and bulk data enrichment.
- AI Query Interface: Natural language → SQL generation
- Multi-workspace: Federated data across environments
- Bulk Enrichment: Excel/CSV processing at scale
VOC & Buyer Journey Chatbot
Production-ready chatbot with dual-brain architecture: SQL Brain (Genie API for structured queries) and Docs Brain (RAG for unstructured PDFs). LangGraph orchestration with LangSmith observability.
- Dual-Brain Architecture: SQL + Document retrieval
- LangGraph State Management: Complex workflow routing
- Vector Search: Databricks BGE-large embeddings
Analytics Data Pipeline
Automated data export pipeline using GA4 Data API. Bypasses UI sampling limits to extract 100K+ rows of unsampled user behavior data for accurate cohort analysis and funnel metrics.
- Unsampled Export: 100K+ rows without sampling
- Cohort Analysis: D1/D7/D30 retention tracking
- Automated Pipeline: Daily scheduled extraction
LinkedIn Campaign Measurement
Automated daily campaign stats export via LinkedIn Marketing API. Company name resolution and comprehensive Excel reporting for cross-platform ad measurement and partner data delivery.
- Daily Automation: Scheduled campaign data export
- Company Resolution: URN → name mapping
- Cross-Platform: Multi-channel attribution reporting
Personal Portfolio
OpenClaw - Personal Productivity Agent
Personal family assistant for deep research. Synthesizes sparse information across news, emails, and market signals into transparent discovery workflows. Automates morning briefs and cron jobs for high-velocity decision making.
- Research: Deep synthesis of sparse info
- Briefs: Cross-channel morning updates
- Agency: Independent cron job execution
Infinite Memory: Scaling Agentic Memory to 10M Turns
Discovery of the 'Discovery Cliff' in LLM memory scaling. Introduces Recursive Gated Consolidation (RGC) to achieve 100% signal recall at extreme scale ($10^7$ turns), maintaining O(1) memory scaling for production-grade AI agents.
- The Discovery Cliff: Identification of 17% recall collapse at scale.
- Recursive Gated Consolidation: Decoupling discovery from history depth.
- Hardware Grounding: TPU v4 OCS and SparseCore optimization analysis.
- Empirical Calibration: Cross-model validation (Gemini 2.5 - 3.1).
Personal AI Infrastructure (PAI v2)
A persistent, self-improving AI Operating System layer integrated across all local projects. Implements standard orchestration cycles and long-term memory via the Kai System for architecture-level persistence and session resilience.
- Orchestration: The Six Cycles of agentic workflow
- Shadow Profile: Secure execution via non-native rituals
- Kai System: Automated decision & learning persistence
- Session Resilience: 429/413 loop awareness & state preservation
Systemic Multi-Agent Coordination
Beyond Chatbots: A dual-layer architecture separating tactical execution from strategic oversight to ensure production-grade AI reliability.
- System 1 (Tactical): Persona-driven sub-agents executing high-velocity task loops (The "Inner Loop").
- System 2 (Strategic): Parent orchestrator enforcing quality gates and architectural safety (The "Outer Loop").
- Synchronized Reasoning: Real-time verification ensuring autonomous execution never breaks project intent.
Autonomous Trading System (V4.1)
Production-grade autonomous trading system managing a dual-track portfolio of Stocks and Crypto. The V4.1 'Steady Winner' strategy achieves 80.8% XIRR (60.3% Alpha over S&P 500) through dual-track asset allocation, Golden Cross momentum, and real-time Shadow Portfolio tracking.
- 80.8% XIRR (60.3% Alpha vs S&P 500)
- Dual-Horizon: Technical momentum vs. long-term fundamentals
- Shadow Tracking: Real-time statistical verification of trade signals
- Tax Efficient: Optimized execution for long-term capital gains
IDE-Agnostic Agent Orchestrator
An orchestration framework that unifies AI agent workflows across VS Code, JetBrains, and CLI. Decouples safety policies (runtime locks, circuit breakers) from specific IDE adapters.
- Policy-First: Mandatory safety and quality gates
- Tool Agnostic: Works across multi-IDE environments
- Circuit Breakers: Prevents infinite agent loops
Astrological Research Platform
Full-stack platform combining astronomical calculations (Swiss Ephemeris) with AI interpretations. Backed by rigorous statistical research with FDR-corrected significance.
- 5,885 figures + 239 countries analyzed
- FDR-corrected significance (q < 0.001)
- 14 yoga types and 6 dosha types validated
- Next.js + FastAPI + PostgreSQL
Philosophy Sage
AI-powered platform for ancient philosophy texts with GraphRAG architecture. Neo4j knowledge graph for contextual retrieval across the Mahabharata, Bhagavad Gita, and other classical texts. Sanskrit TTS integration.
- GraphRAG: Knowledge graph + vector search hybrid
- Neo4j: Entity relationships across 18 Parvas
- Sanskrit TTS: Audio playback of original shlokas
- LangChain: Conversational retrieval chains
Sonic Geometry Visualizer
Interactive exploration of Music Theory × Physics × Math. Real-time oscilloscope, spectrum analyzer, Lissajous figures, Circle of Fifths, and polyphonic virtual keyboard—all native Web Audio API.
- Real-time FFT: Spectrum analyzer + oscilloscope
- Lissajous Curves: Harmonic ratio visualization
- Circle of Fifths: Interactive music theory lab
- Zero Dependencies: Pure vanilla JavaScript
Technical Stack
Experimentation & Statistical Modeling
A/B Testing (Design & Analysis), Causal Inference, Hypothesis Testing, Power Analysis, Regression, Forecasting, Python (SciPy/StatsModels).
AI & Machine Learning
LLM Integration (LangChain/LangGraph), RAG Pipelines, Text-to-SQL, Embeddings, Vector Search, Model Serving.
Metrics & Data Infrastructure
Metrics Design, Data Pipelines, SQL/BigQuery/Databricks, ETL Automation, Reporting Tools, Marketplace Analytics.