Marshall Rasendria Mahendra

BISINDO sign language platform with a two-way translator and a 3D avatar teacher

• All camera inference runs on-device, so no video ever leaves the browser.
• 26-class alphabet classifier at 99.31 percent validation accuracy, plus a 32-class word-level 1D CNN at 83.13 percent signer-independent accuracy running at 15 fps with 6 ms median latency.
• Rigged 3D avatar performs sign poses derived from dataset landmarks via analytic inverse kinematics.

Next.js /TypeScript /TensorFlow.js /MediaPipe /react-three-fiber

AI-powered deepfake detection to stop media-based fraud

• Trained a custom deepfake detection model from scratch on 10,000+ images and videos, reaching 90+ percent detection accuracy.
• Built the entire web application independently, end to end.
• Delivered as a social impact initiative educating 40+ residents in a rural village about AI-generated media fraud.

PyTorch /OpenCV /Next.js

API intelligence platform that maps your entire codebase in seconds

• Auto-detects routes, request bodies, and backend-to-frontend contract mismatches directly from source code, with zero annotations required.
• Parses entire codebases and generates a full API map in seconds, aimed at teams with undocumented or out-of-sync APIs.
• In active development with open feedback loops from peer engineering teams.

TypeScript /RAG /Regular Expressions

Open-source plugin that turns AI coding assistants into a senior ML research engineer

• Auto-routes a plain-language task to the best method across computer vision, medical imaging, NLP, tabular, time-series, and more.
• Encodes a research-first, leakage-aware workflow grounded in 2021 to 2026 literature, with realistic-ceiling estimation and split audits.
• Ships an iterative accuracy-improvement loop with per-problem metric and explainability selection.

Claude Code /Antigravity /23 skills /14 domains

Benign vs malignant classification on CBIS-DDSM, done honestly

• My leakage-free approach produced the best results in a 3-person team and was selected as the final pipeline.
• Benchmarked 6 architectures including a custom CBAM hybrid-fusion model and Mammo-CLIP across 12+ experiments.
• Clinical metadata proved the biggest lever, AUC 0.80 to 0.86, verified with Grad-CAM evidence of correct lesion focus.

PyTorch /CBIS-DDSM /Grad-CAM /IEEE

Document automation for 50+ student organizations at Telkom University

• Generates official Proposal Kegiatan and LPJ documents without the chaos of manual formatting.
• Delivered in a self-imposed 3-month sprint with full creative freedom and no predefined specification.
• Solved multi-page PDF rendering with embedded tables, signatures, logos, and pixel-accurate formatting, then passed a penetration test from Telkom's PuTI Directorate.

PHP /dompdf /MySQL

Controlled Comparison of Multimodal Hybrid Deep Learning for Benign-Malignant Mammogram Classification on CBIS-DDSM

JADIMI: Journal of AI-Driven Informatics and Management Information

Vol. 1 No. 1, August 2026, pp 20-33 · First author

Marshall Rasendria Mahendra, Farhan Muamar Fawwaz, Naisya Aghis Nabila, Muhammad Rakha

Many deep learning studies on CBIS-DDSM report accuracy above 95 percent, yet they partition the archive at image level, so views of one patient land in both the training and the test set and the figures cannot be trusted. This paper rebuilds the comparison...more