John Zolton

Meet Murph

About Me

Patent attorney turned software engineer with a passion for building innovative AI-powered tools.

I heard AI was going to take my job so I'm doing my part to speed it up.

When I'm not at the computer or staring into the abyss, I'm usually hiking a trail with my dog, Murph.

Professional Experience

ML Engineer

Sept 2025 – Present

Trunk Tools

•Launched new Submittal Register product by productionizing NLP research into a Prefect-orchestrated AWS pipeline, delivering an app-ready service end to end
•Led cross-team integration by defining data contracts and coordinating API interfaces between NLP and frontend/app teams, unblocking delivery and aligning pipeline outputs with product requirements
•Drove major latency improvements through parallel execution and pipeline tuning, processing ~3,000-page documents in ~5–8 minutes (~4× faster than early runs, beating <10-minute target)
•Implemented token-efficient consolidation and deduplication across parallelized paths, reducing redundant LLM work while preserving deterministic outputs
•Improved extraction quality through evaluation-driven iteration, achieving ~99.6% IoU on page-range detection and higher section-splitting and data-extraction accuracy
•Optimized source-text highlighting by constraining bounding-box search to predicted pages with ±2-page fallback (capturing ~99.8% of misses), reducing local runtime from 10–15 minutes to ~4–5 minutes while maintaining highlight coverage
•Added production robustness for LLM-dependent steps with retry handling and token-range chunking to prevent request failures and improve pipeline reliability

Python

Prefect

AWS

LLM

NLP

Pipeline Orchestration

Performance Optimization

ML Architect

Jan 2025 – Sept 2025

Skyward IT Solutions

•Transformed organization's #1 most popular tool (CMS chat) from rigid path-following to flexible agentic system using Model Context Protocol and custom workflows
•Expanded agent capabilities by building MCP integrations with key tools (Slack, Jira, Confluence, Linear, Internet, Code Execution, Vector and SQL DBs)
•Enabled local spreadsheet analysis with LLMs via client-side code execution and tailored Python-specific coding agents
•Built specialized research agent with custom tooling and prompts to amplify internal contract procurement team, with MCP tools for accessing proprietary data
•Developed and tested multiple knowledge base approaches (vectorstores, agentic-SQL queries, BM25 searches) to identify optimal solutions for specific use cases
•Developed Terraform-based infrastructure-as-code solution for reproducible LLM inference engine benchmarking across cloud providers
•Created custom performance testing tools capturing LLM-specific metrics (time-to-first-token, throughput) for 4 inference engines and 25+ open-source models
•Onboarded and trained interns and new hires in developing effective agents with the MCP protocol and best practices for LLM-powered applications

Python

AWS

LLM

MCP

Llama-Index

Terraform

Knowledge Engineering

BM25

Vector Databases

Cloud Infrastructure

Performance Testing

Founder/Creator

Jan 2025 – Present

Socratic Sensei

•Developing AI-powered study platform for law students using Next.js, tRPC, and LLMs
•Built spaced repetition system that automatically generates personalized issue-spotting essays based on students' knowledge gaps
•Implemented knowledge-grounded LLM architecture that anchors AI responses to students' course notes, eliminating hallucinations while providing Socratic guidance

Next.js

tRPC

LLMs

Spaced Repetition

Knowledge-grounded AI

Founder/Creator

Jan 2024 – Aug 2024

Patense.ai

•Developed Patense.ai using Next.js, tRPC, Prisma, Tailwind, and AWS Lambda to help patent lawyers analyze documents
•Built cloud and privacy-first local solutions utilizing LLMs, with 2X performance boost through vLLM implementation
•Increased search speed by 300% via prompt engineering and created scalable data processing with custom O(nlog(n)) algorithm searches

Next.js

tRPC

Prisma

Tailwind

AWS Lambda

LLMs

vLLM

Prompt Engineering

Algorithm Design