Backend and AI engineer specializing in production RAG pipelines, tool-calling agents with memory, and multi-tenant backends with scalable AWS infrastructure. At APS Data Technologies I've shipped systems serving the 7th largest Illinois K-12 district and 5 university partners — owning backend architecture, deployment, and the AI features end-to-end.
I build things that hold up in production, not just in demos.
Career
Experience
April 2025 — Present
Software Engineer
APS Data Technologies — EdTech B2B
Building production RAG pipelines, tool-calling AI agents with per-user memory, and multi-tenant backends with scalable AWS infrastructure serving the 7th largest Illinois K-12 district and 5 university partners. Owned backend architecture end-to-end for async processing, vector search, and a capability-based RBAC system districts use to manage their own roles and permissions at runtime.
September 2024 — April 2025
Software Engineer
Safety Straw — Seed-stage Safety Tech Startup
Built an AI support agent with custom tools and sub-agents that resolved 80%+ of customer inquiries autonomously. Shipped the REST API backend with Stripe integration and async order workflows, plus the React storefront end-to-end as the sole engineer.
September 2024 — May 2025
Operations & Growth Lead
Imagine Software — Student Organization
Scaled student tech org 55% from 450 to 700 members. Conducted 20+ technical interviews for client project staffing.
Aug 2022 — May 2026
B.S. Computer Science, Business Minor
Michigan State University
GPA 3.75. SpartaHack XI Blockchain Track 3rd Place (among 100+ teams).
Case Studies
Selected Work
Production Work
ERP Platform for K12 Schools
2025 — Present
Multi-tenant ERP connecting HR, payroll, and staff provisioning for K-12 school districts. Replaces manual clock-in, paper-based hiring workflows, and IT-gated access changes. Serves 1,650+ daily active users at the 7th largest Illinois school district.
Backend Engineer @ APS Data Technologies
Django RESTCeleryNext.js 15PostgreSQLAWSTerraform
Problem
Districts ran HR, payroll, and staff provisioning across disconnected vendor portals with no unified view of a candidate or employee's status, no self-service role management, and manual candidate screening that took 2+ hours per batch.
What I Built
Async hiring pipeline (Django + Celery + Redis) with idempotent retry and a dead-letter queue that keeps screening off the API request path so a Redis blip doesn't silently drop a candidate.
Capability-based RBAC (deny-by-default, module.submodule.action) with a central permission registry that materializes on deploy and purges stale grants automatically, so districts manage roles through the admin UI without filing a ticket or triggering a redeploy.
Tool-calling AI agent connected to live background-check and payroll APIs with per-user session memory, surfaced through a Next.js 15 frontend with DocuSign offer signing and SSO.
Impact
1,650+ DAU at the 7th largest Illinois K-12 district; screening cut from 2 hours to 12 minutes; AI agent reduced routine ERP task time 5× by consolidating three separate vendor portals into one interface.
Admissions & Career Platform for University Partners
2024 — 2025
AI-powered student support platform embedded with 5 university partners across the US and India. Guides international students through US college applications, internship searches, hackathons, and financing. RAG pipeline with cross-encoder reranking and per-user memory handles 200+ daily queries at 95%+ retrieval accuracy.
International students applying to US colleges had no advisor coverage across time zones and no way to get answers grounded in their own documents (essays, transcripts, visa situation) rather than generic guidance.
What I Built
Async OCR pipeline that parses each student's uploaded documents on intake and embeds them into Qdrant, so every retrieval query is scoped to that student's actual materials, not a shared knowledge base.
Two-stage RAG pipeline (OpenAI embeddings → cross-encoder reranker) with semantic chunking aligned to document structure, where the reranker closes the gap when the right document isn't the most similar one to the raw query.
Per-user memory layer (Postgres + LangGraph) tracking document metadata and session summaries so returning students pick up where they left off without re-explaining their situation each session.
Impact
5 university partners onboarded; 200+ daily queries at 95%+ retrieval accuracy; OCR pipeline replaced manual per-applicant document review for every new upload.
Full-stack build for a seed-stage startup making straws that chemically glow when a drink has been spiked. Owned the website, Stripe-integrated checkout, async notification system, and an AI agent for B2B order management.
A seed-stage startup needed a working e-commerce system for two distinct channels (individual B2C purchases and bulk B2B venue orders) before launch, with no dedicated support staff to handle the inquiry volume that comes with it.
What I Built
REST API (Node.js + Express + MongoDB) with JWT auth, Stripe checkout sessions, and async post-purchase notification queues that decouple payment processing from notification delivery so a slow email provider doesn't stall the checkout response.
Separate B2B bulk order flow with volume pricing and restock logic isolated from B2C checkout, since mixing the two channels would have made both worse and complicated the agent's order lookup surface.
Tool-calling AI support agent with sub-agents that queries live order and shipment data through internal and third-party APIs, escalating cleanly when a query is outside tool coverage rather than guessing.
Impact
AI agent resolved 80%+ of customer inquiries autonomously at a company with zero dedicated support staff; shipped 25% ahead of schedule as the sole engineer across backend, frontend, payments, and AI.
Other Projects
Multi-Agent Simulation Engine
2026
Real-time civilization simulation supporting up to 50,000 concurrent agents. The scheduler is world-agnostic: any simulation plugs in by implementing a single DoAction interface, and the scheduler stays completely blind to agent and action types. Scaled using a 4-tier priority system (critical/gameplay/environment/misc) with async job queues, separating deterministic combat from construction tasks to prevent frame stalls.
Product Lead — Team of 30
Scheduling Architecture
Implemented stride scheduling where processes track virtual time position, with stride inversely proportional to priority. The scheduler always advances the furthest-behind process, ensuring fair proportional CPU allocation across 50,000+ agents.
Organized execution into 4 tiers (CRITICAL 40%, GAMEPLAY 30%, ENVIRONMENT 20%, MISC 10%) with independent Scheduler instances per tier via composition.
Set per-tier budgets with soft enforcement for CRITICAL (250ms overage allowed) and hard cutoff for others, preventing lower-tier processes from starving critical gameplay loops.
Built in C++23 compiled to WebAssembly via Emscripten, designing clean interfaces between world state, agent behaviors, and rendering systems for a distributed 30-person team.
Verified correctness across unit tests (Scheduler add/remove/priority/peek), integration tests (budget enforcement, tier isolation, edge cases), and starvation scenarios.
Demo showing three production patterns: capability-based RBAC (why deny-by-default scales better), two-stage RAG (embedding + cross-encoder tradeoff), and agentic systems with tool execution and memory.
Demo Project — open source
RBAC: Districts need self-service role management — capability-based, deny-by-default, so a missing grant means denied without maintaining deny rules.
RAG: Support agent needs 95%+ accuracy at sustainable cost — two-stage pipeline uses embedding for recall, cross-encoder for precision.
Agent: Resume screening must complete even if client disconnects — Celery task, not streaming HTTP.
Django RESTNext.js 14PostgreSQLRedisCeleryQdrantClaude API
Course review aggregator where only verified university students (via school email) can submit reviews. Tracks 5 rating dimensions (overall, difficulty, materials, workload, fairness) per professor and semester.
Lead Developer (85 of 153 commits)
Tracks 5 independent rating axes (overall, difficulty, materials, workload, fairness) with granularity per professor and semester.
Implemented pagination and incremental data fetching so the UI stays responsive as review count grows, avoiding full dataset loads on mount.
Used Supabase RLS to enforce per-student data isolation, handle school email domain validation, and provide OAuth without custom auth infrastructure.
Volunteer matching system using composite scoring across proximity, skills, urgency, and reliability. Integrated Ethereum smart contracts for on-chain donation tracking. Won SpartaHack XI Blockchain Track 3rd Place.
Team of 4 — SpartaHack XI | Blockchain Track 3rd Place
Scoring algorithm weights proximity (geodistance), skill overlap, job urgency (time-decay for stale requests), and volunteer reliability rating to produce ranked matches.
Built real-time messaging between matched volunteer and requester without exposing personal contact info, using JWT-authenticated channels.
Deployed VolunteerLeaderboard smart contract on Ethereum for transparent donation tracking, integrated to frontend via Ethers.js.
