01 NVIDIA Inception Application

TailorMeSwiftly

AI career infrastructure for job seekers and institutions. 30+ tools, three engines, one founder. Seeking NVIDIA NIM integration for document intelligence at the parsing layer.

Company
Tailored Services LLC
Founder
Steve Szopinski
Location
Orchard Park, NY
Live at
tailormeswiftly.com
02 The Problem

Resumes are parsed by machines,
but written for humans

75% of resumes are rejected by ATS before a human sees them. Job seekers don't know why. Career centers can't scale 1-on-1 coaching. The gap between what applicants write and what systems read is a document intelligence problem.

75%
Resumes filtered by ATS
before human review
40+
ATS platforms with different
parsing behaviors
7 sec
Average recruiter
resume scan time
03 The Product

Three engines. 30+ tools. Shipped.

Apply Engine

ATS-optimized resume tailoring with semantic keyword matching, recruiter heatmap simulation, and multi-format export. Supports Workday, Greenhouse, Lever, iCIMS, Taleo, and 40+ ATS platforms.

Learn Engine

Adaptive skill development using spaced retrieval practice (Ebbinghaus, Leitner, Pimsleur). Skill gap analysis against O*NET/ESCO taxonomies. Mastery-based progression, not time-based.

Inform Engine

AI-curated industry briefings with multi-voice podcast generation. Source credibility scoring, recency weighting, and transparent curation logic. No black-box algorithms.

30+
Career tools shipped
32
Edge functions in production
96
Test files (unit + E2E)
04 Current Architecture

Privacy-first, serverless, AI-native

Resumes parse locally in the browser before any data reaches an AI model. Generation runs through Supabase Edge Functions proxying Gemini. Deployed on Vercel with GitHub Actions CI/CD.

Client (Browser)
PDF.js / Mammoth.js
Local resume parsing
Vanilla JS App
30+ tools, 3 engines
Backend (Supabase)
Edge Functions (Deno)
32 serverless functions
PostgreSQL + Auth
RLS, user profiles
AI Layer
Google Gemini API
Generation & tailoring
NVIDIA NIM
Document intelligence (planned)
05 NVIDIA NIM Integration Plan

Where NIM fits:
document intelligence at the parsing layer

Our generative pipeline (Gemini) handles text generation. NIM handles the structured understanding that generative models aren't optimized for.

1. Layout-Aware Document Understanding

Use NIM vision models to analyze resume visual structure — section boundaries, column layouts, header hierarchies. Powers our recruiter heatmap simulation with actual document layout data instead of heuristic guessing.

  • Visual structure extraction from PDF renders
  • Section classification (experience, education, skills)
  • Reading-order inference for multi-column layouts

2. Named Entity Recognition for Skill Extraction

Use NIM NER models to extract structured skill, role, and experience entities from unstructured resume text. Enables semantic matching against job descriptions at the entity level, not string level.

  • Structured skill/role/certification extraction
  • Entity-level job description matching
  • O*NET/ESCO taxonomy alignment

Why NIM, not Gemini for this?

Generative LLMs hallucinate structure. When we ask Gemini to extract skills, it invents ones that aren't there. NIM inference-optimized models give us deterministic, structured extraction — the right tool for the parsing layer, leaving generation to the generative model.

06 Market

Two channels. One platform.

B2C — Job Seekers

Direct consumer SaaS at $9.99/mo. 30+ tools accessible immediately. Free tier with premium upgrade. Chrome extension for in-browser tailoring.

B2B — Institutions

Annual campus licenses for university career centers, community colleges, workforce development programs. FERPA-ready, WIOA-aligned, LTI 1.3 integration for Canvas/Blackboard/Moodle/D2L. Institutional admin dashboard with cohort management and outcomes reporting.

Institutional alignment

Eligible for DOL Strengthening Community Colleges grants, Perkins V, Title III/V. Career centers can include TMS in federal proposals as a technology partner.

07 Differentiators

What separates TailorMeSwiftly

  • Privacy-first architecture — resumes parse locally before reaching any AI
  • No fabrication principle — AI tailors what you wrote, never invents experience
  • Three engines, not one tool — Apply + Learn + Inform as an integrated platform
  • 40+ ATS platform coverage — not generic "ATS optimization"
  • Institutional-grade — LTI 1.3, FERPA, cohort management, outcomes reporting
  • Solo founder velocity — 30+ tools, 32 edge functions, 96 test files, mobile app, all shipped by one builder
  • Research-backed methods — spaced retrieval, cognitive load optimization, eye-tracking heatmap research
  • Evidence over claims — every feature is live, not a roadmap slide
08 NIM Integration Roadmap

Implementation timeline

Month 1–2

NIM NER Integration — Skill Extraction Pipeline

Deploy NIM NER endpoint via build.nvidia.com. Build extraction pipeline in Supabase Edge Function. Replace heuristic keyword matching with entity-level skill extraction. Validate against O*NET taxonomy.

Month 3–4

NIM Document Understanding — Layout Analysis

Integrate NIM vision/OCR models for resume layout analysis. Feed structured layout data into recruiter heatmap simulation. Support multi-column, non-standard, and designed resumes that PDF.js text extraction misses.

Month 5–6

Institutional Deployment & Benchmarking

Deploy NIM-powered pipeline to institutional partners. Benchmark extraction accuracy vs. baseline. Publish results for NSF SBIR preliminary data. Scale inference for campus-wide deployments.

09 The Ask

What we need from NVIDIA Inception

NIM API Credits

Access to NIM microservices for document understanding and NER models via build.nvidia.com. Prototyping and production inference for resume parsing pipeline.

Cloud Credits

GPU-backed compute for NIM container deployment. Scale inference for institutional campus deployments serving thousands of students per semester.

Technical Guidance

Access to NVIDIA engineering support for optimizing NIM model selection and deployment architecture for our document intelligence use case.


Website
tailormeswiftly.com
Research
tailormeswiftly.com/research
Institutions
tailormeswiftly.com/institutions
Founder
Steve Szopinski