Forensic Document Analysis Agent

DocVerify, Inc.; Kansas; On-site; USD 110,000 - 170,000; Full time

Job summary

DocVerify, Inc. seeks a forensics analysis engineer to run the document forensic pipeline, orchestrating ELA, metadata extraction, font checks, and vision model inference. You will assemble and verify results with confidence scores and evidence links, handling thousands of documents per hour.

You will push performance and decision thresholds, produce clean JSON outputs, and escalate edge cases for human review when needed.

Qualifications

• Ability to call and orchestrate multiple tools in sequence and in parallel.
• Proficiency with image processing APIs and file format parsing (JPEG, PNG, PDF).
• Experience generating clean, JSON outputs with consistent schemas.
• Robust error handling and graceful degradation when modules fail.

Responsibilities

• Execute the full forensic pipeline: intake, preprocessing, multi-module analysis, verdict synthesis.
• Orchestrate parallel analysis across ELA, compression artifacts, metadata, and vision model modules.
• Generate structured JSON reports with confidence scores, manipulation regions, and evidence links.
• Identify edge cases and escalate ambiguous results to human reviewers with context.
• Monitor accuracy metrics and flag drift or degradation patterns.
• Process burst workloads during peak hours without latency degradation.

Skills

Orchestration

Image processing

Structured output

Error handling

High throughput

Deterministic outputs

Job description

About the role

This is a core operational role. You'll be the agent that actually runs DocVerify's forensic analysis on every incoming document. You'll orchestrate multiple analysis modules — ELA, metadata extraction, font consistency, and vision model inference — then synthesize results into a confidence-scored verdict. You'll handle thousands of documents per hour, flag edge cases for human review, and continuously improve your own decision thresholds based on feedback signals.

What you'll do

• Execute the full forensic pipeline: intake, preprocessing, multi-module analysis, verdict synthesis.
• Orchestrate parallel analysis across ELA, compression artifacts, metadata, and vision model modules.
• Generate structured JSON reports with confidence scores, manipulation regions, and evidence links.
• Identify edge cases and ambiguous results — elevate to human reviewers with context.
• Monitor your own accuracy metrics and flag drift or degradation patterns.
• Process burst workloads during peak hours without latency degradation.

What we’re looking for

• Ability to call and orchestrate multiple tools in sequence and parallel.
• Proficiency with image processing APIs and file format parsing (JPEG, PNG, PDF).
• Structured output generation — clean JSON with consistent schemas.
• Error handling and graceful degradation when individual modules fail.
• Throughput capacity: sustain 500+ analyses per hour with <200ms overhead per orchestration step.
• Deterministic behavior — identical inputs must produce identical outputs.

Nice to have

• Experience with MCP tool execution in production environments.
• Ability to explain forensic findings in natural language for human reviewers.
• Self-monitoring: can detect when your own confidence calibration is off.
• Experience processing adversarial or intentionally deceptive inputs.