Jaydip Pokiya

AI Researcher

York, Ontario, Canada11 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Innovative AI researcher with strong mathematical background.
Expertise in NLP and computer vision applications.
Proven track record in developing automated pipelines.

Stackforce AI infers this person is a Machine Learning Engineer with expertise in AI-driven video processing and aerospace applications.

Contact

Skills

Core Skills

Natural Language Processing (nlp)Deep LearningOptimization

Other Skills

Python (Programming Language)C++MATLABSOLIDWORKSResearchDevelopment (R&D)Reinforcement LearningDistributed TrainingImage ProcessingStable DiffusionDistillationDeep Convolutional Generative Adversarial Networks (DCGAN)Information RetrievalRetrieval-Augmented Generation (RAG)Document Processing

About

I like to innovate and implement Machine Learning algorithms to solve meaningful problems. I have strong Mathematical and Statistical background and aptitude for problems solving. I am constantly learning and staying up to date with the latest research, especially in NLP, 2D/3D computer vision and vLLMs. I enjoy exploring innovative ways to devise new algorithms and architectures. I’m driven by the challenge of connecting cutting-edge research with practical applications to create meaningful solutions.

Experience

11 mos

Total Experience

11 mos

Average Tenure

Current Experience

University of toronto - rotman school of management

Research Assistant

Oct 2023 – Jul 2024 · 9 mos · Toronto, Ontario, Canada · Remote

Developed a TikTok live stream scraping utility using Python.
Performed speech-to-text conversion using Whisper and compared the performance of Roberta family models
on sentiment analysis task to select the best fit.
Developed a fully automated end-to-end pipeline for scraping data, performing sentence-wise sentiment
analysis, and displaying results on video.

Natural Language Processing (NLP)Python (Programming Language)

Amd

Video Engineer

Jun 2021 – Dec 2022 · 1 yr 6 mos · Hyderabad, Telangana, India · Remote

AI-enabled Content-Aware Bitrate Control Policy for AMD MA35D Accelerator
Developed a comprehensive software pipeline over the AV1 video encoder including Deep learning for content classification, a control policy for the rate controller, and hand-crafted metrics for performance evaluation.
Processed and analyzed encoder look-ahead histograms to construct quality datasets. Created a frame complexity classification module and designed rate control policy, achieving ∼25% additional bitrate saving.
Used HOG for face detection and developed a Region of interest-based PSNR and SSIM metrics.
Built an automated inference pipeline using C++, data structures and OpenMP. Optimized the pipeline
by leveraging model quantization to deliver a scalable high-performance POC.

Deep LearningC++

Indian institute of science (iisc)

Research Assistant

Aug 2019 – Jul 2020 · 11 mos · Bangalore Urban, Karnataka, India

Devised a computationally efficient and fast converging optimal control algorithm ”Constrained Model Predictive Static Programming (C-MPSP)”. The algorithm solves a nonlinear optimization problem with state and control as "path constraints" while satisfying the terminal boundary conditions as “hard constraints”.
Simulated a terminal phase of a 3D nonlinear missile guidance problem with lead angle, lateral acceleration, and impact angle constraints using C-MPSP.
Performed critical analysis study of one of the previously developed techniques ”Unscented Model Predictive Static Programming” by simulating angle-constrained missile guidance that minimizes remaining target estimation errors.

MATLABOptimization

Csir-national aerospace laboratories

Research Intern

Jan 2018 – Jun 2018 · 5 mos · Bangalore, India.

Created a mathematical model for loss estimation across a turbine stage using MS Excel automation.
Carried out aerodynamic design and structural analysis of a single-stage highly loaded transonic axial turbine and studied the effects of three different vortex distributions on the turbine designs.
Generated performance plots by processing experimental data from the turbine test rig.

SOLIDWORKSResearchDevelopment (R&D)