Zeke Lares
I build the systems that turn the physical world into data — at scale, securely, and increasingly with AI in the loop. Fifteen years from first commit to running all of engineering.
I'm Ezequiel "Zeke" Lares, Head of Software Engineering at Butlr, where I lead backend, IoT, and infrastructure. I set technical direction for the org while staying close enough to the code to be dangerous.
My path runs from being the first corporate developer at an engineering firm to scaling an API by several orders of magnitude. Along the way: secure VPN backends at Norton, spatial data ontologies at Mapped, and a sensing platform at Butlr that understands buildings without ever pointing a camera at anyone.
A lot of my work lives where engineering meets AI: putting LLMs into real production paths, not demos. Off the clock, I run an open source cancer research project and wrote a book about it.
From BI reports and internal tooling to global IoT fleets and APIs at internet scale: each role one step deeper into systems and leadership.
A cross-literature synthesis that maps where cancer-therapy research is concentrated, where the gaps are real versus artifacts, and which mechanistic ideas deserve broader comparison — then uses Monte Carlo and spatial simulations as a claim-testing engine. Open-source and free forever.
A simulation library written in Rust with Python bindings (MIT-licensed). The simulations bound the conditions under which a hypothesis could hold — they're computational predictions, not modality verdicts. Manuscript complete; headed to Amazon.
Foundational infrastructure written from scratch in Rust — durable, crash-safe, and frugal by design. Each project uses AI to mine prior art and adversarially verify every load-bearing claim before trusting it.
A durable, crash-safe message queue for the edge — in one static Rust binary.
One ordered queue that survives power loss and self-heals against corrupt files with bounded, reported loss. SQS-style visibility leases, group-commit fdatasync durability, and zero-copy fan-out — drops onto a Raspberry Pi with no JVM and no dependencies. Verified with deterministic crash-injection simulation.
The most efficient Redis-compatible cache — in one static Rust binary.
A shared-nothing, thread-per-core engine that speaks RESP2/RESP3 and keeps the Redis contract — built for maximal throughput per core and minimal memory per item. Borrows the best from Dragonfly, Valkey, and the modern eviction literature (W-TinyLFU, S3-FIFO, SIEVE), with forkless snapshots.
A modern, standards first OpenID Connect identity platform.
Multi tenant, multi environment, EdDSA first, built toward a certified OIDC core with the conformance suite wired into CI as a merge gate. One repository holds the provider server, management APIs, SDKs, admin SPA, and hosted auth pages. Single static binary on PostgreSQL.
A traffic and API manager in one Rust binary.
An L4 and L7 load balancer, reverse proxy, API gateway, and Kubernetes ingress controller in a single binary, with a dashboard that can actually change things. Zero downtime for both configuration and binary swaps: atomic snapshot apply plus descriptor handoff, connections preserved across each.
Open-source, AI-driven cancer-therapy research — and the engine behind the book.
A cross-literature synthesis of 4,830 papers plus a Monte Carlo and spatial simulation library (MIT-licensed Rust, with Python bindings) that bounds the conditions under which mechanistic claims could hold. The codebase behind the manuscript — free forever.
Butlr AI Platform for Private Spatial Intelligence: anonymous people sensing that uses body heat and AI to understand presence, movement, and occupancy without ever capturing an identity.
Won by the Butlr platform, where I lead engineering. See the winner listing.
I'm really used to working by myself. Now, working with a partner, listening to his ideas — it's more creative and stimulates my ideas, too.
— Zeke, featured as an undergraduate researcher in "Computer Design A Big Hit", on the NSF Research Experience for Undergraduates program.
Research abstract: Spring 2009 ECS abstract booklet (PDF)
Scaling stories, AI in production, and the occasional detour into open science.
A dozen people and about a hundred sensors when I joined. Thousands of sensors around the world today. The road between, as my commit history remembers it.
ReadShipping LLMs into real production systems — not demos. What it took to trust a model enough to put it on the hot path.
ReadA side project that turned into 4,830 papers, a simulation engine, and a book. Here's what building it taught me.
ReadOpen to conversations about engineering leadership, AI in production, and the cancer-research project. The fastest way to reach me is email.