Position : Senior Python Developer
Experience : 5+ Years
Shift : 12 PM – 9 PM IST
Location : Remote (India)
About the Company
We are the world’s first independent auditing body dedicated to verifying human authorship across creative assets. Our mission is to distinguish authentic human work from AI-generated content for agencies, publishers, and enterprise brands.
We conduct forensic audits across video, image, text, and code to secure intellectual property integrity.
Role Overview
We are building a proprietary Internal Forensic Dashboard —a local, browser-based tool that analyzes multi-modal assets using advanced algorithms to detect signatures of AI generation.
This is a strictly NO-AI-CODE environment.
No ChatGPT, CoPilot, Cursor, or LLMs may be used to generate code.
We audit codebases, and any synthetic structures result in termination.
We want real engineers who understand math, computer vision, and Python internals —not API callers.
Key Responsibilities
🔹 1. Video Forensics (Highest Priority)
- Implement FFmpeg-based frame extraction and keyframe sampling
- Analyze temporal consistency , pixel noise, and lighting variance
- Build Deepfake Detection using Error Level Analysis (ELA)
- Perform frame-by-frame artifact scanning
🔹 2. Image Forensics
Implement Error Level Analysis (ELA) using OpenCVBuild FFT frequency-domain analysis to detect GAN / Diffusion patternsExtract Metadata, EXIF, C2PA tags🔹 3. Text Forensics
Integrate Hugging Face models (e.g., RoBERTa) for :Perplexity scoringBurstiness measurementImplement stylometric analysis (sentence variance, structure patterns)🔹 4. Code Forensics
Build Git history “ velocity analysis ”Parse AST Structures to detect AI-generated code blocksIdentify non-human coding patterns🔹 5. UI + Dashboard Development
Build a local browser-based UI using Streamlit or FastAPIEnable asset uploads for all four modalitiesDisplay visual outputs (heatmaps, FFT plots, graphs)🔹 6. Forensic Report Generator (Legal-Grade PDF)
Build dynamic reporting engine using Jinja2 + ReportLab / FPDF / WeasyPrintEmbed forensic visuals programmaticallyAdd auditor overrides & human-in-the-loop notesToggle evidence sections on / offRequired Skills
🔥 Technical Must-Haves
Strong Python engineering experience (5+ years)Deep understanding of :NumPyOpenCVFFmpegComputer vision mathExperience processing large video & image files locallyStrong foundational math : luminance analysis, gradient detection, spectral analysisExperience with Streamlit, FastAPI or similar frameworksPDF generation experience using :ReportLabWeasyPrintFPDFStrong debugging & system-level reasoning💡 Mindset Requirements
Must write clean, human-authored, verifiable codeNo AI-generated code allowedMust provide GitHub linksComfortable working independentlyNice to Have
Background in security forensicsExperience with C2PAExperience with feature extraction from multimedia assetsFamiliarity with deepfake datasets