Artificial Intelligence Engineer

Contract Senior Full-Stack AI Engineer

Recruiting Intelligence Platform — United Global Technologies (UGT

)
About U

GTUnited Global Technologies is a Charlotte, NC–based IT staffing and solutions firm serving federal and SLED clients nationwide. We're a six-time Inc. 5000 honoree with WOSB, SBE, DBE, and MBE certifications. We move fast, we know our market, and we're investing in building proprietary technology that gives our recruiters an edge no off-the-shelf ATS provide

s.
We're building an internal recruiting intelligence platform from the ground up and need a senior engineer to bring it to life. Today our resumes live in a legacy ATS that can only do keyword matching. We're replacing that with a system our recruiters log into to do something no tool on the market does for us: search our entire candidate library — 300,000 resumes and growing — using plain-English criteria or a full job description, get an instantly ranked shortlist with the reasoning behind each match, and launch compliant, personalized outreach to those candidates by text and ema

il.
You'll be the engineer who builds it. We have a clickable prototype of the recruiter experience and a documented architecture; we need someone who can turn that into real, production software our team uses every day. This is a build-from-scratch role with real ownership, not a maintenance

gig.
What you'll build an

• d ownDocument ingestion at scale. Our resumes come out of the old system as documents only — PDF and Word, no structured data. You'll build the pipeline that parses them (including scanned/OCR cases), extracts a clean structured profile from each using an LLM, and handles 300k in a one-time backfill plus a steady incremental flow as new resumes ar
• rive.Semantic search. Embeddings + a vector database powering a two-stage search: cheap retrieval narrows the full library to a strong shortlist, then an LLM scores and explains only that shortlist. Search must stay fast as the library g
• rows.Candidate de-duplication / identity resolution across hundreds of thousands of records with no reliable IDs (matching on extracted email plus fuzzy fallb
• ack).The recruiter web application — search, ranked results with match reasoning, candidate profiles, and an outreach workflow — with authentication and role-based ac
• cess.Outreach integration. RingCentral SMS (high-volume API, with opt-out/consent respected) and an email service, with per-recipient personalization and a human review-and-approve step before anything s
• ends.A security and compliance posture appropriate for candidate PII under federal contracting — access controls, audit logging of screening decisions, and outreach that respects TCPA/CAN-SPAM. You don't need to be a compliance lawyer, but you must build like someone who takes data protection serio

usly.
You're a strong fit if yo

• u haveSenior-level full-stack experience shipping production web applications end to end (front end, back end, database, deploy
• ment).Hands-on experience building LLM / GenAI applications — specifically RAG, embeddings, and vector search (e.g., pgvector, Pinecone, Weaviate, or Qdrant) using a major model API (Anthropic, OpenAI, or sim
• ilar).Experience with data/ingestion pipelines and document parsing (PDF/Word,
• OCR).Comfort integrating third-party APIs (SMS, email, or similar) into a real wor
• kflow.A track record of owning architecture decisions and working directly with a non-technical stakeholder — translating business goals into technical choices and explaining trade-offs in plain lan
• guage.A security-conscious approach to handling sensitive personal

data.
Nice

• to haveExperience with HR tech, ATS systems, or recruiting wor
• kflows.Exposure to federal contracting, OFCCP/EEO, or other regulated-data enviro
• nments.Prior RingCentral (or comparable telephony/SMS API) integ
• ration.Experience scaling search or data systems into the hundreds of thousands or millions of r

ecords.
How we

'll workYou'll work directly with Jason as product owner. He's non-technical, so clear communication, regular demos of working software, and the ability to make and defend sound architecture decisions independently are essential. We'd rather see working increments early and often than a big reveal at

the end.Deliverables / mi

• lestonesPhase 1 (primary): Ingest the full resume library and ship the search experience — plain-English criteria and job-description matching, with ranked, explained
• results.Phase 2: The outreach workflow (RingCentral text + email, personalized, with the approve-before-sen
• d gate).We'll agree on a milestone schedule and acceptance criteria u

p front.