525Krequests/month on production ML infrastructure
150klines → agent-native codebase
10xquery latency reduction (4s → 0.4s)
300kresume chunks searchable in <1s via RAG
466safe merge requests shipped without regressions
2wks→2dprototype cycle acceleration via agent workflows
Summary
Founder of Syntropy Systems, building Elwood, an AI worker, and open-sourcing the tooling that tests agents and schedules their experiments: windtunnel (agent reliability bench), pit (a prediction market that schedules experiments), and whirr (local experiment orchestration). 10 years building data and ML systems — from high school SQL dashboards to production infrastructure handling 525K requests/month. Interpretability on consumer hardware: a sparse-autoencoder replication of Golden Gate Claude on Qwen2.5-1.5B and a backward "ghost-token" Jacobian lens on Qwen3-0.6B. Research background: 70-paper literature synthesis on AI safety, backpropagation from the 1986 Rumelhart paper, research mentorship designing paper curricula. Architected sub-1s semantic search across 300k documents. Created the Agent Tokens protocol. Built 5-agent orchestration pipeline handling 150k+ line edits with zero regressions. WPI Data Science graduate (3.75 GPA, High Distinction). BlueDot AI Safety Alignment certified.
Building Elwood, an AI worker that gives teams leverage over the tools they already use, and asks for a go-ahead before doing anything new
Open-sourced the agent testing and experimentation stack: windtunnel (reliability bench for tool-using agents; 19 releases, on PyPI as windtunnel-bench), pit (a prediction market that schedules experiments), and whirr (local experiment orchestration)
Independent interpretability research on consumer hardware: replicated Anthropic’s Golden Gate Claude with a sparse autoencoder on Qwen2.5-1.5B, and extended Anthropic’s Jacobian lens with a backward “ghost-token” lens
May 2025 - Dec 2025
Chief AI Officer · CazVid LLC (Bootstrapped, ~20 employees)
Design AI systems using frontier models; evaluate tradeoffs between approaches (cost, latency, accuracy) and select architectures that match business constraints
Transformed 150k-line legacy codebase into agent-native code: 1400 lint fixes → 900 type resolutions → 300-test suite. Agentic coding now works like greenfield.
Founded Palmera Labs (internal R&D): RAG systems, agent frameworks, fast-weight plasticity research
Technical leadership: onboard 7-engineer team on AI coding agents, translate CEO/COO requirements into technical roadmaps
Drove 466 merge requests with zero production regressions via automated guardrails
Sep 2024 - May 2025
Director of AI · CazVid LLC
Nashua, NH
Achieved sub-1s full-text search across 300k resume chunks via LLM/RAG + MongoDB hybrid filters
Architected Kern ML service: FastAPI intake → Kafka queue → Redis state → NestJS/Pusher delivery
Built universal matching algorithm using triplet loss embeddings + hybrid vector/text search
Created no-code interface enabling non-engineers to deploy Codex-generated branches to production
Synced GitLab CI/CD with GitHub Actions enabling repo-level context for AI agents
Engineered AWS infrastructure (ECS, S3) supporting 10k+ DAU with auto-scaling
Architected web platform extending mobile app, increasing user engagement 40%
Led post-acquisition technical integration: unified CI/CD, consolidated 3 codebases into 1
Sep 2021 - Jun 2022
Full Stack Engineer / Junior Data Scientist · Agency Leads
Manchester, NH
Built Vue.js + Express.js data entry platform from scratch, processing 50k+ records
Reduced Typesense query latency from 4s to 0.4s (10x improvement) via index tuning
Increased data production 125% by implementing work classification specializations
Deployed AWS ECS infrastructure with Terraform enabling zero-downtime deployments
Created Tableau dashboards driving $2M+ in data-informed sales decisions
Sep 2020 - May 2021
AI/ML Capstone Project (MQP) · Dell EMC
Worcester, MA
Built predictive model for hardware failure detection using deep learning on log data
Applied Chi-Squared + Pearson correlation for feature selection across 500+ variables
Presented weekly to Dell engineers; model deployed for internal diagnostic testing
Early Experience
Jun 2020 - Aug 2020
Research Intern · WPI Computer Science Dept
Worcester, MA
Built COVID-19 demographic dataset aggregating UN, WHO, and Johns Hopkins data. Wrote Python pipelines, hosted on MySQL, published to Kaggle for research community.
Aug 2017 - Jan 2020 (2.5 years)
Software Engineering Intern · CyberSN
Boston, MA
Full stack development on internal sales tooling (Blitz). Rapid prototyping, documented codebase architecture. Attended DEF CON 2019 (company-sponsored).
May 2015 - Jul 2017 (2 years)
Data Intern · Biscom
Westford, MA
First technical role at age 16. SQL queries on production databases, built user analytics dashboards in Excel, presented findings to company president and engineering team.
AI Safety Compass — 70-paper literature review exploring model meta-alignment. Original survey methodology mapping where models sit on alignment/openness axes. [Paper]
Backprop Paper Replication — Hand-derived gradients from Rumelhart et al. 1986, implemented on Apple MLX. No AI assistance — just the paper and framework docs. [GitHub]
Introducing Agent Tokens — Open standard for AI agent identity and intent: signed, JWT-style tokens that declare what an agent is allowed to do. [Website]
Project EVE: Novel Launchsondes for Atmospheric Data — IEEE Xplore published. Co-author on environmental sensor network paper. [Paper]
Research Mentorship
Summer 2025 — Designed and led intern research project investigating bias in embedding-based hiring systems.
View the paper curriculum I designed
Language Models are Few-Shot Learners (GPT-3)
Chain-of-Thought Prompting Elicits Reasoning
ReAct: Synergizing Reasoning and Acting
Sparks of AGI: Early Experiments with GPT-4
Core Views on AI Safety: When, Why, What, How
Constitutional AI: Harmlessness from AI Feedback
The Superintelligent Will (Bostrom 2012)
Man is to Computer as Woman is to Homemaker? Debiasing Embeddings
Synthetic Data Generation: A Survey
Demystifying Embedding Spaces using LLMs
What are polysemantic neurons?
Methodology: Structured research board (hypotheses → experiments → paper draft), weekly syncs, "stay one week ahead" cadence.
Output: Unpublished paper applying utilitarian frameworks to LLM ethics, investigating gender bias in synthetic resume retrieval systems.
Open Source Projects
Wind Tunnel — Reliability bench for tool-using LLM agents — reproducible scenarios scoring outcome, trajectory, and constraints, with an experiment ladder that gates regressions. 19 releases; on PyPI as windtunnel-bench. Open source from Syntropy Systems. [GitHub]
Pit — A prediction market that schedules experiments: agents fund runs and bet PASS/FAIL, and the runs they disagree about go first. Append-only git ledger. Open source from Syntropy Systems. [GitHub]
whirr — Local experiment orchestration: SQLite job queue, one worker per GPU, metrics dashboard. On PyPI. [GitHub]
jacobian-lens — Backward "ghost-token" extension of Anthropic's global-workspace Jacobian lens, with causal patch tests on Qwen3-0.6B. [GitHub]
Golden Gate Qwen — Replication of Anthropic's Golden Gate Claude on consumer hardware. Sparse autoencoders on Qwen2.5-1.5B for interpretable feature discovery and steering. [GitHub]
Agent Tokens Protocol — Signed, JWT-style tokens that declare an agent's identity and allowed scope. [Website]
Resume MCP — Model Context Protocol server exposing identity as queryable API. Deployed on Cloudflare Workers. [Live]
Crystallize — Python framework for reproducible ML experiments. On PyPI as crystallize-ml. [GitHub]
DeltaTask — MCP server for Obsidian task management. SQLite-backed. [GitHub]
p5.interact — p5.js 2.x addon: hovered(), clicked(), dragged(), dropped(), scoped by push/pop. Published on npm. [Docs]
gcomm — Rust CLI that writes Git commit messages with a local LLM via Ollama. Published on crates.io. [crates.io]
Leadership
Founder, Palmera Labs — Internal AI R&D division at CazVid (2025)
Founder, Syntropy Systems, Inc. — AI workers and open-source agent tooling (2024-Present)
President, Phi Sigma Kappa — WPI Chapter, 50+ members (2019)
Secretary, SIAM — Society of Industrial and Applied Mathematics (2018)
Status: Building Syntropy full time · open to research collaborations Based in Nashua, NH | US Work Authorized