Denis Timonin Email & Phone Number

San Francisco, Ca, Us

FlyMy.AI – the ultimate inference platform for all your AI needs

Us

Santa Clara, Ca, Us

Contributor and AI Solutions Architect of the GPT-3 training, inference, parameter-efficient tuning pipelines in the NeMo-Megatron project. Contributor to the FasterTransformer framework ( that enables the world's fastest transformer inference). Co-author of the pioneering course on Huge Neural Networks (GPT).

Santa Clara, Ca, Us

As NVIDIA’s first AI Architect for the CIS region, I led deep learning research, CUDA programming, neural network optimization, full-stack AI architecture, and consulted major enterprises across 12 countries.Achievements: • Developed AI approaches presented at AI conferences, including: • Framework for GPT-3 inference on multiple GPUs, achieving ~2x acceleration • Accelerated GAN training in Mixed-Precision, resulting in ~2x faster StarGANv2 training • Built around 30 AI pipelines for major CIS enterprises in Telecoms, Finance, Retail, Robotics, Call Centers, and Social Networks • Worked on training the largest NLP GPT model in the region • Contributed to popular AI frameworks, including: • TensorRT compiler, enabling ~60x speedup of UBER’s CoordConv layer • 10x acceleration of Object detection pipeline with TRITON • 4x speedup of GPU-based Vector search

Shenzhen, Guangdong, Cn

Neural Network developed and trained by me was integrated into HUAWEI mobile phones.Research, training, and inference for small, fast, and precise neural networks for mobile devices.

AI startup. Snapchat for real goods (try-on real glasses and jewelry using your mobile phone)Founder and Lead Researcher at Ideofit.Me and Coingauge.io.- Built from scratch and led R&D team of 8 developers; - Succesfully built three AI projects: CV on mobile phones (ideofit.me), ASR, NLP for analyzing twits.- Organized R&D process for team of Machine Learning developers;- Developed DL/CV algorithms for mobile application Ideofit.me (face/facial keypoints detection, face size detection, credit card detection).- Implemented ML `anti-fraud` system for payment gateway, number of `fraud` transactions were reduced on ~80%.- Implemented Neural Network for emotion detection on smartphones- Developed mobile neural network for sounds detection on mobiledevices- Implemented client-server application architecture for appliccationhttps://ideofit.me/

Founder at ​bloonic.net​ (IoT. Wi-Fi stand-alone button. Analog of Amazon's Dash Button and Flic):- Staffed a team of interested experts and made MVP version of the product.- Developed all software parts of the project

New York, New York, Us

- Developed an ensemble of algorithms (Computer Vision / Deep Learning) for the segmentation of medical ​MRT/CT​data, as a result: time needed for segmentation of 1 MRT scan were reduced from 16 to 2 hours. Algorithms in process of patenting in USA;- Implemented client-server application architecture for application;- Implemented client-server application for data processing;

San Francisco, California, Us

- “Upwork’s Top-Rated Freelancer” grade. - Handstack.com,​ I optimized code for sms-send `workers`, as a result:system productivity increased x50 times (sms-send-rate increased from5000 to 250000 messages per hour).- Contributor at ​Storeys.​ Realized `url dispatcher` for open sourceweb-framework `storeys`.

- Created ML algorithm to prevent `fraud` withdraw transactions - Developed and integrated: payment gateway, KYC (know your customer) system, TFAsystem, messages sending system

- Built and lead the team of three programmers;- As a Team Lead developed more than 10 web projects:Social network Loove.com, API backends for iOS applications, and more