← Elijah Croft

Elijah Croft

Education

California State University, Fullerton
Fullerton, CA
B.S. Computer Science & B.A. Philosophy (Double Major)
Expected May 2028
  • Undergraduate research in AI ethics; presented a panel on fairness, accountability, and bias in algorithmic decision-making.

Experience

Software Engineer (Contract)
May 2025 – Present
2E4ME
Newport Beach, CA
  • Shipped the full-stack scheduling platform now coordinating 20+ instructors and 70+ students, retiring a manual spreadsheet process that cost staff hours every term; React 19 + TypeScript on an Express REST API.
  • Built the scheduling core as an OptaPlanner constraint solver, resolving availability, room capacity, and enrollment into conflict-free timetables in seconds, on multi-tenant infrastructure secured with Google OAuth 2.0 SSO.
Robotics Coach & Programming Instructor
Mar 2025 – Present
Magikid Robotics Lab
Tustin, CA
  • Led a team of 10 to two VEX AI World Championship finishes (18th and 8th against international fields) and a first-place win at the VEX AI International Open at UC Berkeley, owning code review and autonomous strategy.
  • Drove the autonomous stack: a Markov Decision Process policy cutting path-planning errors 80%, fed by YOLOv11 detection.

Projects

Mini Motorways Route Optimizer|Python, NumPy, Pillow|GitHub
2026
  • Modeled house-to-store assignment as a min-cost flow (successive shortest paths, Johnson potentials) under per-store capacity caps, building the cost matrix in S Dijkstra sweeps instead of H×S — provably optimal, not greedy.
  • Routed roads with a Steiner-forest heuristic over free existing tarmac, then rip-up-and-reroute and reassignment passes to a fixed point: 42–84% fewer road tiles, traffic-shaded so chokepoints surface; a screenshot-to-board vision pipeline re-plans the live game.
Agent Poker — LLM Agents Under Imperfect Information|Python, unittest — no runtime dependencies|GitHub
2026
  • Reproducible harness for LLM agents under imperfect information: no-limit Hold'em with side pots, plus Kuhn poker for its exact equilibrium.
  • Scored policies by exact best-response search over the game tree, not simulation; 182 tests.
evosim — Individual-Based Evolution Simulator|Python, NumPy, pytest
2026
  • Wrote a 7,700-line agent-based simulation in which each organism is a discrete individual with a diploid genome, resolving energy, foraging, predation, and reproduction per tick across a world of depth-attenuated light, thermal gradients, and nutrients.
  • Detected speciation automatically and recorded full phylogeny; runs reproduce bit-for-bit from (config, seed) under 19 test modules.
DormDash — Autonomous Dorm Delivery Robot|C, Python, Zephyr RTOS, ROS2, Raspberry Pi 5, Zenoh, CAD|GitHub
Fall 2025
  • Autonomous hallway delivery robot: a ROS2 graph on a Raspberry Pi 5 over Zephyr RTOS firmware in C, closed-loop PID motor control.
  • Bridged MCU and Pi over Zenoh, designed the 12V Li-ion power path and CAD chassis, and shipped a Next.js delivery queue.
Headless Session Monitor & Alert Service|Python, Playwright, GraphQL, Raspberry Pi, systemd
Jul 2026
  • Reverse-engineered a private GraphQL API from live traffic; a systemd unit on a Pi 4 polls it and pushes mobile alerts.
IQDraw|Python, GitHub Actions|GitHub·Docs
Aug 2026
  • Generated step-by-step VEX IQ build booklets from a text spec: parts at hole coordinates in, isometric step diagrams out.
PPP Simulator|Best Use of ArcGIS, FullyHacks 2026|Next.js, TypeScript, ArcGIS SDK|Demo·GitHub
Apr 2026
  • Browser ocean-drift simulator advecting spills through real ArcGIS ocean-current vector fields, with marine-hazard overlays.
MixSafe|Next.js, TypeScript, Tailwind CSS, Django|Live·Design
Oct 2024 – Present
  • Built and deployed a harm-reduction platform serving 500+ weekly users and ~2,000 monthly interactions, on a structured interaction database backing real-time lookup of substance combinations and their effects.
  • Designed the launch campaign solo — logo system, 23 campus flyers, and 12 rounds of press-ready stickers — each sheet carrying its own QR code, so every scan is attributable to the poster that earned it.

Technical Skills

Languages: Python, C++, TypeScript/JavaScript, Java, C, Scala, SQL
Frameworks: React, Next.js, Node/Express, Django, Tailwind CSS, pytest
ML & Modeling: NumPy, YOLOv11, Markov Decision Processes, agent-based simulation, combinatorial optimization
Systems & Embedded: ROS2, LiDAR SLAM & localization, Raspberry Pi 4/5, Jetson Orin Nano, Li-ion power & BMS, CAD, VEX V5/IQ
Tools: Git, Docker, Linux, Bash, Make, GitHub Actions, Playwright, REST & GraphQL APIs