cv

Below lies my brief CV

Basics

Name Yash Jayswal
Label Software Engineer
Email yashmjayswal@gmail.com
Phone +82 010 7408 2354
Url https://YashJayswal24.github.io/portfolio/
Summary Software Engineer at Samsung HQ, Suwon, South Korea. IIT Delhi graduate (B.Tech Electrical Engineering). Experienced in ASR models, C++ cloud modules, machine learning research, and ML optimisation.

Work

  • 2024.08 - Present
    Software Engineer
    Samsung HQ
    Working on multilingual ASR models and scalable C++ cloud infrastructure for Bixby.
    • Developed an ASR model (31 languages) achieving ≈ 45% latency reduction from baseline.
    • Implemented C++ modules within multi-user Bixby Cloud for Fast Conformer models.
    • Improved beam search module memory usage by 10%.
    • Coordinated global R&D centers for 2025 ASR deployment across 12 languages.
    • Designed robust data collection pipelines using yt-dlp and Python.
  • 2023.05 - 2023.07
    Software Engineer Intern
    Samsung HQ
    Developed phonetic similarity metrics and database modules for speech recognition.
    • Developed Name Entity Replacer for Personal Database module.
    • Proposed novel phoneme similarity metric based on Metaphone and Soundex.
    • Achieved 96.3% and 98.3% accuracy for spoken name matching.

Education

  • 2020.11 - 2024.08

    New Delhi, India

    B.Tech
    Indian Institute of Technology Delhi
    Electrical Engineering
    • Probability & stochastic processes
    • Digital image processing
    • Machine intelligence & learning
    • Data structures & algorithm
    • Graph theory
    • Computer architecture

Awards

Certificates

ResNet Badge
Deep-ML 2026-01-27
Attention Is All You Need
Deep-ML 2026-01-27
Deep Learning Specialization
Coursera / deeplearning.ai 2022-01-13

Skills

Programming Languages
Python
C
C++
Java
Technologies
Tensorflow
Keras
Numpy
PyTorch
MATLAB
jax
Git
Kubernetes

Languages

English
Fluent
Hindi
Native
Gujarati
Native

Projects

  • LLM Scaling & Systems Optimization (JAX/TPU)
    Engineered distributed Transformer components in JAX with TP, PP, and FSDP to overcome auto-compiler scaling bottlenecks.
    • Hand-coded a Mixture of Experts (MoE) layer via jax.shard_map with explicit All-to-All token routing, completely eliminating native AllGather memory explosions.
    • Leveraged XLA trace profiling to hide interconnect latency behind compute, achieving a 43x MoE speedup (73ms to 1.7ms) on Kaggle TPU v5e-8.
  • Semantic Communication
    Developed a generative AI-based semantic communication pipeline targeting 6G communication.
    • Guided Stable Diffusion + ControlNet approach
    • Achieved 0.31 LPIPS score on VGG net
  • Panic-Driven Crowd Dynamics
    Developed strategies to reduce death count in dense crowd panic situations through simulation.
    • Reduced death rate by 70%
    • Applied numerical mollification and optimized obstacle shapes
  • DSCoin - Cryptocurrency
    Built cryptocurrency system using SHA256 hashing and implemented mechanism to validate transactions.
    • Proof-of-work Blockchain in Java
    • Tree-structured Blockchain & broadcasting