AI Agent Systems for
Enterprise-Grade Audit & Trust

Why This Experience Maps to Fieldguide

Direct alignment between Fieldguide's Senior AI Engineer requirements and my track record building agentic systems, evaluation frameworks, and enterprise-grade AI in production.

Enterprise Clients & Domain

Successful B2B deliveries across enterprise, healthcare, insurance, and government clients. Currently building LLM-based AI agents at Pevo, with prior NLP-based chatbot delivery to Hyundai Motor Group, Amore Pacific, Gangnam District Office, and DA Plastic Surgery.

Cresoty (크레소티)

15,000+

Pharmacies served — 15K out of total 25K in the Korean market (~60% domestic market share). Market leader in pharmacy POS systems. Deployed 2 AI agents (document + voice) with complex workflow automation.

LLM-based AI Agent

KB Insurance (KB 손해보험)

POC → Prod

One of South Korea's largest insurers. Built multi-agent AI architecture for automated insurance claim damage assessment with human-in-the-loop verification — the kind of complex, judgment-heavy workflow Fieldguide automates.

LLM-based AI Agent

Hyundai Motors (Hyundai AutoEver)

B2B

Built and delivered AI chatbot system for Hyundai Motor Group via their IT subsidiary Hyundai AutoEver. Enterprise-grade conversational flow with end-to-end delivery.

NLP-based Chatbot

Amore Pacific (Etude)

B2B

AI-powered cosmetics recommendation chatbot for Etude House, a global beauty brand under Amore Pacific — South Korea's largest cosmetics conglomerate.

NLP-based Chatbot

Enterprise AI in Production — Before Pevo

Left: Amore Pacific (Etude House) — AI-powered lip color recommendation via image recognition, deployed to 10,000+ users. Right: Gangnam District Office — Public service chatbot handling citizen inquiries.

IGC 2016 Conference — Presenting "AI Platforms for Game Customer Service". Text-to-vector transformation for automated inquiry classification using CNN.

Three AI Agents, One Platform

Each agent addresses a distinct operational challenge — from document processing to real-time voice interaction — connected through shared OCR/LLM infrastructure. The same pattern of automating complex, judgment-heavy workflows that Fieldguide brings to audit and assurance.

Agent 01 — Insurance

Automated Damage Assessment

Multi-agent system that automates the labor-intensive process of reviewing medical receipts and determining insurance coverage eligibility — a complex workflow requiring structured data extraction, domain-specific rules, and human judgment. Directly analogous to the audit workflows Fieldguide automates.

Data Extraction Pipeline — Vision AI + Medical AI agents digitize handwritten receipts, then RAG agents cross-reference against insurance contracts for coverage eligibility

Data Extraction Pipeline

OCR Module → LLM structured extraction. Handles receipts, ID cards, pet registration docs, and bank account documents. Vision AI + Medical AI agents digitize messy handwritten receipts into normalized JSON.

Coverage Analysis Engine

Strategy pattern per insurer (KB, DB, Meritz, Samsung, Hyundai). Common classification agent (1 LLM call) → per-insurer rule calculation (0 LLM calls). 9-category exclusion framework with surgical procedure linking.

Human-in-the-Loop Interface — Certified loss adjusters verify AI classifications via drag-and-drop, with automated data logging creating a continuous feedback loop for model improvement

Human-in-the-Loop Verification

Certified loss adjusters review AI classifications via drag-and-drop interface. Override capability with structured reason codes. Every manual decision is logged — the same HITL design pattern that audits require, where AI assists but professionals make final judgment calls.

Evaluation & Optimization

Reduced LLM calls from N (one per insurer) to 1 via common classification layer. Individual strategy calculations are pure rule-based — no LLM, no latency. DB-driven rules allow A/B testing without code changes. Feedback loop analytics for continuous model improvement.

Receipt Image → OCR Module → Text Lines → LLM Extraction → Structured Items
                                                              ↓
                                               Common Classification Agent  ← 1 LLM call
                                                    ↓              ↓
                                          covered_items    excluded_items
                                           ↓         ↓         ↓         ↓
                                      KB Strategy  DB Strategy  Meritz  Samsung  ← 0 LLM calls
                                           ↓
                                   Loss Adjuster Review → Approve / Reject / Override
                                           ↓
                                   Feedback Loop → Model Retraining Data

Agent 02 — Pharmacy

Automated Diabetes Prescription Processing

Extracts structured data from 20+ hospital-specific diabetes prescription formats using the same OCR + LLM pipeline — then automatically submits reports to South Korea's National Health Insurance Service (NHIS). Demonstrates the same document intelligence and compliance automation Fieldguide brings to audit evidence collection.

Prescription Processing Pipeline — Vision AI extracts 30+ fields from 20+ hospital-specific formats, with automated reporting to the National Health Insurance Service

Multi-Format Extraction

Auto-detects 3 prescription types (general diabetes, medical aid, CGM continuous monitoring). Type-specific LLM prompts extract 30+ fields: patient info, diagnosis codes (E10-E14), medication items, dosage schedules, insulin usage, and institutional codes.

Compliance Automation

Extracted patient info + diagnosis codes automatically formatted and submitted to the National Health Insurance Service. Dynamic protocol adaptation based on diabetes type and level. End-to-end automation from paper prescription to government submission — the same regulatory compliance rigor audit firms need.

Prescription Image → OCR Module → Text → Type Auto-Detection
                                                    ↓
                               ┌────────────────────┼────────────────────┐
                         General Diabetes    Medical Aid       CGM Electrode
                               └────────────────────┼────────────────────┘
                                                    ↓
                                         LLM (type-specific prompt)
                                                    ↓
                                         Structured Prescription JSON
                                           ↓                    ↓
                                  NHIS Auto-Report     → Agent 03 (Voice AI)
                                                         medication data feeds
                                                         personalized calls

Agent 03 — Voice AI

Medication Adherence Voice Agent

A real-time voice AI agent that calls patients at scheduled times to verify medication intake, check for side effects, and escalate safety concerns. Built on Agent 02's prescription analysis — the system knows exactly which medications each patient takes, their dosage schedule, and relevant drug interactions.

Launched — March 23, 2026

Voice Agent system officially launched at the 2026 수도권 약사학술제 & 48th PHARM EXPO (Suwon Convention Center). Live demo and booth presentation to pharmacists nationwide.

PHARM EXPO 2026 — 팜페이 AI booth showcasing the Voice Agent system

"퇴근후에도 AI가 환자를 케어합니다" (AI cares for patients even after hours)

Medication Schedule — Time-based dosage tracking from prescriptions via Agent 02

AI Call Settings — User configures call times and medication groups

Key Architecture Decision: Dual-Channel Design

Instead of streaming raw audio to the server, the system performs on-device STT and sends only text over a dedicated text channel. A separate audio channel handles TTS playback. This dual-channel WebSocket design minimizes latency and preserves voice quality — the same architectural trade-off thinking required when designing agent systems for enterprise reliability.

VoIP Call Infrastructure

Server triggers FCM (Android) / APNs VoIP Push (iOS) at scheduled times. CallKit integration presents native phone UI. Cross-platform: iOS uses AVAudioEngine + CallKit; Android uses WebRTC + FCM data messages.

On-Device STT (Multi-Provider)

OpenAI Whisper, ElevenLabs Scribe, Google Cloud, Naver Clova, and native speech recognition. VAD with RMS threshold tuning. 0.7s silence detection triggers transcription; 5s max utterance limit.

LLM Conversation Engine

LLM with dynamic system prompt injected with patient's prescription data. Korean AI pharmacist persona with sliding window memory (last 10 turns). Structured flow: greeting → medication check → side-effect screening → safety escalation → farewell.

Real-Time TTS Streaming

Multi-provider TTS (ElevenLabs, Google Wavenet, Naver Clova). Server generates audio chunks → streams over WebSocket → native audio engine playback. TTS caching strategy: pre-generates responses based on patient profiles for minimal latency.

End-to-End Call Flow

1. Scheduled Trigger

Server checks dosage schedule → sends VoIP push notification at configured time

2. CallKit / Native Ring

Patient sees incoming call from "AI Pharmacist" — standard phone UI, works on lock screen

3. Dual WebSocket Channels

Text channel: STT transcripts + LLM responses. Audio channel: TTS audio chunks. Both over WSS.

4. Dynamic Conversation

AI: "Did you take your Metformin 500mg with lunch?" — generated from Agent 02's prescription data

5. Side-Effect Screening

LLM probes for common side effects based on prescribed drugs

6. Safety Escalation

Risk signal detected → connects patient to pharmacy/hospital. Transcript saved for review.

Dosage Scheduler → FCM / APNs VoIP Push → CallKit (iOS) / FCM (Android)
                                                        ↓
                                               Patient accepts call
                                                        ↓
                                        ┌───── Dual WebSocket ─────┐
                                        │                            │
                                  Text Channel              Audio Channel
                                        │                            │
                            On-device STT → text           TTS audio chunks
                                        ↓                            ↑
                              LLM + Prescription DB   →   ElevenLabs TTS
                              (dynamic prompt injection)
                                        ↓
                              Adherence logged to DB
                              Risk? → Pharmacy / Hospital alert

Full-Stack Architecture Ownership

Designed and operated the complete technical stack — from infrastructure and databases to ML pipelines and mobile client. The same end-to-end ownership Fieldguide values.

Production Results

Why I'm Drawn to Fieldguide

What excites me about Fieldguide is the mission-critical complexity. Building AI agents for audit and advisory isn't about simple automation — it's about navigating judgment-heavy workflows where compliance, accuracy, and trust are non-negotiable. This is exactly the kind of work I've been doing: healthcare compliance automation, insurance claim assessment with multi-insurer business rules, and government reporting. The domain is different, but the engineering challenges are the same.

I'm especially aligned with Fieldguide's approach of shipping daily value. My career has been built on bias to building — moving fast from POC to production, iterating based on real user feedback, and owning outcomes end-to-end. Having already shipped 3 production AI agents with full-stack ownership, I'm ready to bring that same velocity and rigor to Fieldguide's audit intelligence platform.

Fieldguide is opening a new office in Seoul — and I'm based here, deeply connected to the Korean enterprise ecosystem, and ready to help build the team from day one.