About this position
JOB DESCRIPTION
AI Architecture Lead – macOS Forensics
Location: Remote (US Preferred)
Company: SUMURI LLC – Magnolia, Delaware
Reports To: Founder / Director of Software
Product Focus: RECON ITR & RECON LAB (macOS-native forensic tools)
About SUMURI
SUMURI is a Delaware-based digital forensics company specializing in macOS forensic
software and hardware used by law enforcement, military, and corporate investigators
worldwide. Our flagship tools — RECON ITR (imaging & triage) and RECON LAB (analysis &
reporting) — are undergoing a modern Swift-native rebuild designed for Apple Silicon and
long-term AI integration.
We are building the most advanced macOS forensic AI platform in the world.
Position Summary
The AI Architecture Lead will design and oversee the long-term AI and ML architecture for
RECON ITR and RECON LAB, ensuring:
● Native Swift/macOS integration
● Apple Silicon optimization
● Offline AI model execution
● Forensic defensibility
● Scalable feature velocity using AI coding agents
● Strict privacy and security standards
This is not a web AI role.
This is not a prompt-engineering role.
This is a macOS-native forensic AI systems architecture role.
Core Responsibilities
1. AI Architecture Strategy
● Design a long-term AI integration roadmap for RECON LAB and RECON ITR
● Architect modular AI pipelines (OCR, face detection, object detection, CLIP-style
labeling)
● Define standards for pretrained model integration (no custom model training required
initially)
● Ensure deterministic, explainable AI workflows suitable for court testimony
2. macOS & Swift Integration
● Architect AI features using:
○ Swift
○ SwiftUI / AppKit
○ Core ML
○ Metal (if needed)
● Optimize for Apple Silicon (M-series)
● Convert PyTorch / ONNX models into Core ML where appropriate
● Ensure compatibility with macOS notarization and sandboxing requirements
3. AI Coding Agent Management
● Design workflows for:
○ Using LLM coding agents safely
○ Automated code validation pipelines
○ Preventing hallucinated unsafe logic
○ Enforcing architectural consistency
● Build structured AI-assisted development pipelines
● Implement guardrails for secure code generation
4. Forensic Integrity & Defensibility
● Ensure:
○ AI outputs are logged and reproducible
○ Chain of custody is preserved
○ Processing is transparent and reviewable
○ No cloud dependency unless explicitly configured
● Design AI workflows that withstand Daubert/Frye scrutiny
5. Performance & Security
● Architect offline-first inference pipelines
● Ensure no unintended data exfiltration
● Implement sandboxed model execution
● Optimize inference performance for:
○ 16GB, 32GB, 64GB Apple Silicon systems
● Reduce memory overhead in large case processing
6. Leadership
● Lead small AI engineering team
● Review Swift and ML code for production quality
● Mentor developers transitioning from C++/QT to Swift
● Collaborate with external development partners
● Set coding standards and documentation requirements
Required Qualifications
Technical
● 7+ years professional software engineering experience
● 3+ years production Swift development
● Deep experience building macOS native applications
● Experience integrating ML models into native applications
● Experience converting models (PyTorch / ONNX → Core ML)
● Strong understanding of:
○ Apple Silicon architecture
○ Memory optimization
○ Concurrency (GCD, async/await)
○ Security best practices
● Experience managing large codebases
AI / ML Experience
● Experience implementing:
○ Object detection (YOLO-style)
○ OCR pipelines
○ Face detection & embedding comparison
○ CLIP-style zero-shot classification
● Experience deploying pretrained models (not necessarily training them)
● Familiarity with:
○ Core ML
○ ONNX Runtime
○ PyTorch
○ Vision framework
● Understanding of deterministic vs probabilistic outputs
Forensic or High-Security Environment Experience (Preferred)
● Experience in digital forensics
● Experience in cybersecurity
● Experience building tools used in regulated environments
● Understanding of evidentiary handling principles
Nice-to-Have (But Not Required)
● Experience testifying or supporting expert testimony
● Experience building offline AI systems
● C++ interoperability knowledge
● Metal acceleration knowledge
● Experience building CLI forensic tools
● Experience with APFS / macOS internals
What Success Looks Like (12–24 Months)
● RECON LAB has modular AI engine framework
● All AI runs offline by default
● AI coding agents reduce feature development time by 40%+
● No AI-related architectural rewrites required
● Clean Swift-native codebase
● Clear AI audit logging system
● Production-ready model update pipeline
● Competitive advantage over SaaS-only forensic vendors
Compensation
Competitive, based on experience.
Equity discussion possible for exceptional candidates.