Back home
Full case study archive

Selected work.

A detailed review of architectures, prototypes and engineering solutions we've deployed. Each study outlines our methodology from research through execution.

Legal document parser

Next.jsPythonLocal LLMChromaDBDocker

Project context

An internal research initiative to evaluate the viability of localized, air-gapped LLM deployments for processing highly sensitive legal documentation without third-party API exposure.

Business challenge

Traditional legal document parsing relies heavily on manual auditing or cloud-based AI APIs, which pose severe data-privacy risks for confidential agreements.

Research findings

Evaluated several quantized models (Llama 3, Mistral) against state-of-the-art closed models to determine the accuracy of zero-shot entity extraction on legal text.

Solution design

Designed a secure retrieval-augmented generation architecture that runs entirely on local infrastructure, so data never leaves the client environment.

System architecture

Next.js frontend communicating with a containerized Python backend. The backend hosts a quantized local LLM, using ChromaDB for rapid similarity search over vectorized document chunks.

Implementation

A custom chunking algorithm tuned for legal clauses (definitions, terms, signatures) and a React dashboard for attorneys to query documents in plain language.

Outcome

A fully offline document query system that extracts clauses with high accuracy, eliminating third-party data exposure entirely.

Lessons learned

Context-window limits in local models demand extremely precise vector retrieval; tuning chunk overlap was critical to preventing hallucinated clauses.

DeFi operations portal

SolidityReactEthers.jsAWSTypeScript

Project context

A foundational prototype for institutional web3 asset management, addressing the need for secure, cross-chain transaction orchestration.

Business challenge

Handling asynchronous, multi-step blockchain transactions in a user-facing dashboard reliably, without dropped states or misleading UI feedback.

Research findings

Investigated common failure modes in decentralized applications — state desynchronization between the frontend and RPC nodes causes most user-facing errors.

Solution design

Architected a state machine that rigidly enforces transaction stages (pending, mined, confirmed, reverted) and polls multiple RPC nodes for consensus.

System architecture

React SPA on AWS CloudFront, interfacing with Ethereum RPC nodes and custom Solidity smart contracts, using Ethers.js for key management.

Implementation

Strict TypeScript typings for every contract ABI, with hardware-wallet support so private keys stay physically isolated.

Outcome

A robust, institutional-grade portal that gracefully handles network congestion and gives deterministic feedback on complex cross-chain operations.

Lessons learned

A single RPC provider is a single point of failure; an enterprise system needs a load-balanced, multi-provider network architecture.