Open to opportunities

Vinay Soni

@vinaysoni

Message

Independent AI Research Engineer optimizing language modeling pipelines for efficient CPU/GPU performance.

India

Message

What I'm looking for

I’m looking to build evidence-driven language modeling systems—tokenization, context modeling, quantization, and CPU/GPU-efficient training/inference—using iterative experimentation, benchmarking, and performance optimization under real compute/storage constraints.

I’m an Independent AI Research Engineer focused on evidence-driven iteration for language modeling. Over 2+ years, I’ve designed and optimized language modeling pipelines—from tokenization (C++ BPE) and training data strategies to long-context mechanisms and efficient attention formulations—while benchmarking compute and storage trade-offs.

In parallel, I build performance-first implementations with custom CUDA, Triton, and SIMD kernels, and I’ve explored hybrid context architectures and learnable quantization that dynamically allocates precision. Previously at Dell Technologies (7+ years), I delivered backend systems and product improvements, worked on a patent-related dictionary-less compression approach, and upgraded platforms for faster performance—bringing an engineering mindset grounded in measurable outcomes.

Experience

Work history, roles, and key accomplishments

Current

Independent AI Research Engineer

Current

Independent

Apr 2024 - Present (2 years 3 months)

Independently designed and optimized language modeling pipelines, including tokenization (C++ BPE) and context modeling mechanisms for long-context performance under CPU and storage constraints. Built efficient attention formulations and custom GPU/CPU kernels, and implemented a learnable quantization framework and SIMD-accelerated inference engine.

C++BPE Tokenizer Vocabulary Refinement Sub word Pruning Training Data Strategy Global Decay Summaries Trigonometric Context Representation Residual Connections Associative Linear Attention Kernel Approximation CUDA Triton SIMD Register Use Precision Allocation Online Learning CPU Based Language Compression

Software Engineer

Dell Technologies

Jul 2015 - Mar 2024 (8 years 8 months)

Developed and maintained backend services, microservices, stored procedures, and UI components for internal and customer-facing products. Worked on platform performance improvements (page loads under 3 seconds), migrated inventory management infrastructure to Pivotal Cloud Foundry, and built integrations/add-ons for the Order Management System.

Backend Microservices Stored Procedures UI Data Migration Cloud Foundry Performance Optimization Customer Feedback Analysis Data Compression

Software Developer, Intern

RRSC-East, ISRO

Jan 2015 - Jun 2015 (5 months)

Developed a Tea Garden Inventory and Management System using ArcGIS and Flex to support tea plantation management.

ArcGIS Flex Inventory Management Tea Plantation Management

Education

Degrees, certifications, and relevant coursework

Manipal Institute of Technology (MIT), Sikkim

Bachelor of Technology, Computer Science and Engineering

2011 - 2015

Grade: CGPA: 8.0

Activities and societies: Undergraduate coursework included Algorithms, Operating Systems, Databases, Programming Languages, Computer Architecture, Computer Networks, and Computational Theory.

Earned a B.Tech in Computer Science and Engineering (CGPA: 8.0) from Manipal Institute of Technology, Sikkim.