Looking for a job

Mahmoud Mousatat

@mahmoudmousatat

Message

CTO at Nuha Tech, where I built a reconstruction pipeline proven on 50,000+ images across 10 museum-scale sites.

Russian Federation

Message

At Nuha Tech, I took a large-scale inverse-rendering research prototype to a production reconstruction pipeline with approximately 7M RUB per year in commercial traction. I designed a framework combining NeRF, 3D Gaussian Splatting, and diffusion-based enhancement, and made full reconstruction practical on consumer-grade GPUs.

I also built distributed training over fVDB sparse-voxel representations, enabling reconstruction from 50,000+ images across 10 museum-scale sites. For Tatar Miras, I directed a browser-based cultural-heritage experience with WebGL2 rendering and a WebGPU water simulator.

My doctoral research focuses on CAD-centric agentic sparse 3D reconstruction: recovering editable geometry from few views with execution and semantic verification. At Innopolis University, I teach computer vision, artificial intelligence, and information retrieval, and have supervised capstone projects and student startups.

Experience

Work history, roles, and key accomplishments

Current

Instructor

Current

Innopolis University

Aug 2024 - Present (2 years 3 months)

Teach Introduction to Computer Vision, Introduction to Artificial Intelligence and Information Retrieval; lead laboratory sessions for Introduction to Machine Learning and Data Wrangling & Visualisation. Supervised 10 capstone projects and 5 student startups, and was named Best Teaching Assistant by student and faculty vote.

Computer Vision Machine Learning Information Retrieval Teaching Mentorship Data Wrangling Data Visualization

Current

Chief Technology Officer

Current

Nuha Tech

Sep 2023 - Present (3 years 2 months)

Direct the technical programme of a deep-tech startup in large-scale inverse rendering, taking a research prototype to a production pipeline with sustained commercial traction of approximately 7M RUB per year. Designed an end-to-end reconstruction framework integrating NeRF, 3D Gaussian Splatting and diffusion-based enhancement, and led a team of five engineers delivering real-time WebGL2 renderin

NeRFs Gaussian Splats Inverse Rendering FVDB WebGL WebXR ThreeJS React three fiber PlayCanvas Distributed Training GPU Optimization Team Leadership

Data Scientist

Advanced Engineering School

Jun 2023 - Feb 2024 (8 months)

Adapted a 13B-parameter language model to the Haskell programming language using QLoRA (4-bit NF4 quantised backbone with low-rank adapters), fitting the full fine-tuning run onto a single NVIDIA A100. Built the supporting data pipeline end to end: large-scale source-code corpus collection, deduplication, syntactic and semantic validation, and a reproducible training and checkpointing workflow.

QLoRA LoRAs Large Language Models Haskell Data Pipelines A100 Fine Tuning Reproducible Training

Research Intern

Innopolis University

Jun 2023 - Jul 2023 (1 month)

Benchmarked GCN, GAT and GIN architectures against fingerprint baselines on SMILES-encoded molecular datasets using RDKit and DeepChem, across five featurisation schemes. The circular-fingerprint regression pipeline gave the strongest held-out performance, reaching R² = 0.87 and MSE = 0.31.

graph neural networks GCN GAT Gin RDKit DeepChem SMILES Molecular Property Prediction regression

Backend Team Lead

Innopolis University

Jun 2022 - Jul 2022 (1 month)

Led backend development of a multi-user document annotation platform (Flask, SQLAlchemy, REST): authentication, role-based authorisation, annotation storage, validation logic and relational schema design.

Flask SQLAlchemy REST APIs Authentication Authorization Relational database design Team Leadership