profile-pic

Rishabh kandwal

AI engineer skilled in foundation models, LLMs, and generative AI. Experienced in fine-tuning Transformers and building AI-driven automation and POCs on cloud platforms.

  • Role

    AI Engineer

  • Years of Experience

    3.2 years

Skillsets

  • MLOps
  • Data Pipeline
  • Git
  • GPT
  • Gradio
  • HPC
  • HuggingFace
  • Linux
  • LoRA
  • cloud deployment
  • Process Optimization
  • QLoRA
  • rag
  • Scripting
  • Service automation
  • Weights & Biases
  • Byt5
  • OpenCV
  • Azure
  • BART
  • Fairseq
  • Java
  • Keras
  • LangChain
  • Mt5
  • NLTK
  • Python - 2.0 Years
  • PyTorch
  • Scikit-learn
  • spaCy
  • SQL
  • TensorFlow
  • AWS
  • ChromaDB

Professional Summary

3.2Years
  • Jun, 2024 - Present2 yr 1 month

    Project Engineer

    CDACINDIA
  • May, 2022 - Sep, 20231 yr 4 months

    Project Engineer

    Wipro

Work History

3.2Years

Project Engineer

CDACINDIA
Jun, 2024 - Present2 yr 1 month
    Built and deployed high-accuracy Kannada-English transliteration models, hosted on Hugging Face. Fine-tuned Transformer models (ByT5, T5) using LoRA/QLoRA and distributed training. Built an automated LaBSE-based pipeline to align English-Kannada sentence pairs for AI data augmentation. Developed context-aware question-answering pipelines with LLMs and vector databases. Explored LLM-based code refactoring and generation using tools like GitHub Copilot and Hugging Face Transformers. Automated data collection and preprocessing pipelines, generating 100K+ Kannada text samples for model training.

Project Engineer

Wipro
May, 2022 - Sep, 20231 yr 4 months
    Automated ticket classification and log analysis, reducing manual effort. Collaborated with IT Infrastructure teams to identify automation opportunities in support workflows using AI-based tools. Worked with ITSM tools (Jira) and cloud services (AWS, Azure) to streamline ticket resolution and reporting. Enhanced customer query handling systems with NLP, improving response accuracy.

Major Projects

4Projects

KannadaEnglish Transliteration Model

    Built a custom Transformer-based model in TensorFlow with character-level tokenization; trained on CPU with limited dataset and deployed on Hugging Face, achieving performance comparable to state-of-the-art systems.

ByT5 Fine-tuning with LoRA/QLoRA

    Generated high-quality English-Kannada dataset using a hybrid approach and fine-tuned ByT5 with LoRA, QLoRA, and selective layer freezing for improved efficiency and accuracy.

Parallel Corpus Generation using LaBSE

    Created a large-scale English-Kannada parallel corpus using LaBSE for automatic alignment and data augmentation, improving dataset quality for AI model training and automation workflows.

Automated Speech Corpus Generation

    Developed pipeline using OpenAI Whisper ASR to scrape and transcribe Kannada audio from web sources, creating large-scale parallel text-speech datasets for NLP training.

Education

  • Post Graduation Diploma in Advanced Computing

    Centre for Development of Advanced Computing (2024)

  • Bachelor of Technology in Computer Science and Engineering

    Dr. A. P. J. Abdul Kalam Technical University (2022)