What we build.

A Solidity smart-contract system for multi-path token swaps on BNB Chain. Uses golden-ratio path splitting and precision math (Lagrange multipliers, Newton's method) with built-in MEV protection, reentrancy guards and batch operations. Extensively exercised on testnet with static-analysis-driven security hardening.

A real-time, multi-chain arbitrage-detection engine (BSC, Polygon) that analyses on-chain swap events and pool imbalances to surface multi-hop opportunities, with async RPC failover, circuit breakers and REST/WebSocket APIs.

A DeFi analytics platform in .NET for real-time price discovery, arbitrage detection and liquidity monitoring over REST and WebSocket. Built on a clean layered architecture with Nethereum integration and OpenTelemetry observability.

The early C++ graph-model and pathfinding research engine for discovering arbitrage cycles across pools — the foundational R&D that informed our current arbitrage stack.

An automated options-trading platform that executes vertical-spread strategies from signal sources to Interactive Brokers, with real-time P&L, position and risk management, commission tracking and multi-follower copy-trading. Microservices architecture (async Python / FastAPI), OAuth2/JWT security and full CI/CD across unit, integration and end-to-end tests.

A Go service that decomposes numerical and decision outputs into explicit, queryable causal chains with confidence scoring, driver analysis and what-if simulation — explainability infrastructure we apply to trading signals, regime scores and risk metrics for auditable decisioning.

A research system for regime-adaptive trading that combines multi-factor alpha (momentum, low-vol, reversal) with VIX-based market-regime detection and a reinforcement-learning environment, plus a backtest engine with transaction-cost modelling.

A Go framework for coordinating multi-agent AI swarms, implementing distributed-consensus protocols (Byzantine, Raft, Gossip), continual-learning techniques and a federation hub with MCP integration, on a clean domain/application/infrastructure separation.

A local-first orchestration layer for AI coding agents that adds durable state and recovery, multi-machine federation (mTLS) and enterprise governance (SSO/OIDC, hash-chained audit logs, RBAC, OpenTelemetry). Built in TypeScript with a Next.js dashboard and Helm charts.

A multi-agent system that turns documents into presentations through a LangChain + LLM pipeline (extraction, generation, optimization agents) with a FastAPI backend and Celery async orchestration — an example of our agent-orchestration patterns.

The reproducible-research companion to a 2026 academic paper on LLM coding-agent conformance, cost and correctness — including frozen pre-registrations, twelve experimental phases and validation scripts that reproduce its numerical claims with cluster-robust and mixed-effects analysis.