Megh Makwana

Co-Founder

Pune, Maharashtra, India8 yrs 7 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Led India's Sovereign AI revolution at NVIDIA.
Developed state-of-the-art speech models for multiple languages.
Architected large-scale GPU computing clusters for data science.

Stackforce AI infers this person is a Generative AI and Deep Learning expert with extensive experience in large-scale infrastructure and model development.

Contact

Skills

Core Skills

Generative AiSolution ArchitectureAi PlatformsDeep LearningSpeech RecognitionAi DeploymentData ScienceGpu ComputingMachine Learning

Other Skills

EngineeringModel DevelopmentCloud InfrastructureSpeech SynthesisTranslation ModelsTensorRTTriton Inference ServerTensorFlowHorovodGPU Accelerated ComputingAccelerated ComputingNVIDIA Software StackPredictive MaintenanceConvolutional Neural NetworksTechnical Training

About

Applied Generative AI Engineering Manager working at the intersection of building foundational models, optimising workloads across thousands of GPUs and helping CSP build their AI Platforms using NVIDIA AI Enterprise.

Experience

8 yrs 7 mos

Total Experience

2 yrs 1 mo

Average Tenure

6 yrs 2 mos

Current Experience

Nvidia

3 roles

Manager - Applied Generative AI

Promoted

May 2022 – Present · 3 yrs 11 mos

Technical Leader for Solution Architecture & Engineering org at NVIDIA responsible for Sovereign AI, NVIDIA Cloud Partners, Enterprise and Public Sector. I lead a team of generative AI researchers and performance engineers who work at the heart of India's Sovereign AI revolution.
1. Building India's Sovereign Multimodal Foundational Models with our Sovereign LLM Partners.
2. Deploying India's Largest AI Infrastructure with NVIDIA Cloud Partners

Generative AISolution ArchitectureEngineeringAI Platforms

Senior Solutions Architect - Applied Deep Learning

Dec 2021 – Jun 2022 · 6 mos

Key Technical Contribution:
1. Technical Lead for developing Speech Recognition, Speech Synthesis and Translation models for Indic Languages in NVIDIA Nemo/RIVA.
2. Contributed state of the art Speech Recognition models for Hindi, Marathi, Tamil in NVIDIA Nemo.
3. Placed #2 on Interspeech 2022 Gram Vaani Hindi Speech Recognition Open Challenge under the name of NV.
4. Worked with India’s Number 1 Telecom in deploying Question Answering Pipeline on TensorRT and Triton Inference Server improving the throughput by 6x in production.
5. Accelerated Distributed Large Scale Weather Forecasting Application for IITM Pune by 56x using TensorFlow, Horovod.
5. Adjunct Faculty at IIT Kharagpur delivering three microcredits course namely Accelerated Data Science, System Software Engineering for Applied Deep Learning and Hardware/System’s Engineering.
6. Delivered multiple technical sessions at IIT Kharagpur, IIT Jodhpur, IIT Hyderabad, NIT Trichy, National Informatics Centre, IISER Pune, VIT, VJTI, NIT Pondicherry.
Team Contribution:
1. Helped scale the team for Professional Services, HER, Public Sector Applied Deep Learning SA Team.

Speech RecognitionSpeech SynthesisTranslation ModelsTensorRTTriton Inference ServerTensorFlow+2

Solutions Architect - Applied Deep Learning

Feb 2020 – Dec 2021 · 1 yr 10 mos

Top Performer as IC2.
Working with enterprise, government, consumer internet companies in applying the science of GPU accelerated computing for their large scale data science workloads using various GPU accelerated software stack.

GPU Accelerated ComputingData ScienceMachine LearningGPU Computing

Ccs computers private limited

Solutions Architect - Applied Deep Learning

Jun 2018 – Jan 2020 · 1 yr 7 mos · New Delhi Area, India

Worked with a wide range of customers from various segments, helping them architect and deploy large scale accelerated computing clusters for various workloads in data science primarily applied machine learning / deep learning using NVIDIA's accelerated software & hardware stack.
Skilled at using various GPU accelerated platforms & python based numerical computing frameworks like:
1. RAPIDS: Suite of CUDA Enabled Data Science Libraries.
2. TensorRT & ONNX: Deployment of optimized DNN.
3. Horovod: Distributed Computing for DNN coupled with Keras/TF
4. PyTorch: Prototyping & Experimentation of DNN.
5. Keras/Tensorflow: Prototyping & Experimentation of DNN
6. Jax: Experimenting here.
7. DeepStream SDK: Building Video Analytics Pipelines.
Major Contribution:
1. Helped various organization apply state of the art DNN architectures in the space of Computer Vision & Natural Language Processing. Reducing training time and accelerating inference throughput pipelines on various H/W platforms like GPU-based Data Centers i.e DGX/EGX family, GPU-based Embedded Platforms i.e Jetson family.
Problem Statements in Computer Vision:
Classification, Detection, Segmentation, Image to Image Translation using variants of CNN & GANs.
Problem Statements in Natural Language Processing:
Classification using variants of Transformer based models i.e BERT & AWD-LSTM based models i.e ULMFit.
Using various training methodologies like supervised, self-supervised based on problem statement & data availability.
2. Help customers architect and deploy large-scale GPU-based data-center leveraging Docker Platform and orchestrating it using Kubernetes/SLURM.
Have done the largest deployments in India at various prestigious research institutes ranging from 0.5 to 5 Peta-flop of Mixed Precision Compute for large scale machine learning workloads.
3. Authored in few top machine learning conferences/workshops.

Accelerated ComputingMachine LearningDeep LearningNVIDIA Software Stack

Rabbit and tortoise technology solutions

Machine Learning Engineer

Jan 2018 – May 2018 · 4 mos · Pune Area, India

Research and development in Deep Learning
Have worked closely with the team to build state of the art supervised and unsupervised models from international research papers related to predictive maintenance.
Implemented the following research papers:
A) Classification of surface defects on steel sheet using convolutional neural network, achieving accuracy of 97%.
B) Anomaly detection and fault disambiguation in large flight data: a multi modal deep autoencoder approach.
C) Collaborative filtering recommendation system using Amazons data set using deep autoencoder, knn.

Deep LearningPredictive MaintenanceConvolutional Neural NetworksMachine Learning

Softrace systems and solutions

Co-Founder

Jun 2017 – Dec 2017 · 6 mos · Pune

1. Conducted workshops for technical and soft skill training in various colleges like:
a)PDEA College of Engineering, Hadapsar
b)PDEA Polytechnique, Hadapsar
b)JSPM College of Engineering, Hadapsar
2)R&D in machine learning and deep learning.

Technical TrainingMachine Learning