FFFF HARISHKUMAR1112001 (PRAJAPATI HARISHKUMAR KISHORKUMAR) ยท GitHub
[go: up one dir, main page]

Skip to content
View HARISHKUMAR1112001's full-sized avatar

Highlights

  • Pro

Block or report HARISHKUMAR1112001

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please donโ€™t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
HARISHKUMAR1112001/README.md

Hi ๐Ÿ‘‹, I'm HarishKumar Prajapati

Lead Software Engineer | AI/ML Specialist

Passionate about AI Safety & Trustworthy AI systems, specializing in deepfake detection, information retrieval, and LLM fine-tuning. Leading teams to create secure, production-grade AI applications. ๐Ÿš€


๐Ÿ’ผ Professional Experience: 2.5+ Years

Lead Software Engineer & Team Lead at Tricon Infotech (2022-2025)

  • Led engineering teams in developing production-grade AI/ML applications
  • Built intelligent retrieval systems with vector databases and semantic search
  • Architected scalable ML pipelines and LLM integration systems for enterprise applications

MSc in Artificial Intelligence

  • Research focus: Deepfake Detection, Multimodality, AI Safety, Information Retrieval

๐Ÿ› ๏ธ Technical Skills & Technologies

AI/ML & Deep Learning

pytorch tensorflow huggingface scikit-learn pandas numpy

Programming & Development

python javascript nodejs c++ java

Databases & Cloud

mongodb postgresql azure gcp docker

Tools & Frameworks

git linux opencv graphql


๐Ÿš€ What I'm Working On

  • ๐Ÿ”ญ Current Focus: Leading AI/ML teams & developing AI safety systems with focus on deepfake detection
  • ๐ŸŽ“ Studying: MSc in Artificial Intelligence (specializing in AI Safety, Deepfake Detection & Multimodality)
  • ๐Ÿง  Research: Deepfake music detection, neural audio fingerprinting, robust AI systems, LLM fine-tuning
  • ๐ŸŽฏ Exploring: Reinforcement Learning for preference learning (RLHF), democratizing fine-tuning workflows, optimizing pretraining strategies
  • ๐Ÿ’ก Building: Deep learning models for media authentication, information retrieval, and trustworthy AI systems
  • ๐ŸŒฑ Learning: Advanced transformer architectures, adversarial robustness, multimodal learning, AI security
  • ๐Ÿ‘ฏ Collaboration: Open to AI safety research, deepfake detection projects, and secure information retrieval systems

๐Ÿ“š Areas of Expertise

  • AI Safety & Security: Deepfake detection, robust AI systems, adversarial defense, media authentication
  • Information Retrieval: Vector databases, semantic search, neural fingerprinting, similarity matching
  • Large Language Models (LLMs): Fine-tuning, prompt engineering, LoRA/PEFT adapters, GRPO training, RLHF
  • Reinforcement Learning: Preference learning, reward modeling, human feedback optimization
  • Deep Learning: PyTorch, TensorFlow, transformers (BERT, GPT, T5, Qwen), pretraining optimization
  • Audio Processing: Neural audio fingerprinting, music analysis, synthetic audio detection
  • MLOps: Model evaluation, deployment, optimization, production pipelines
  • Team Leadership: Code reviews, mentoring, technical training, project coordination

๐Ÿ”ฌ Recent Projects

๐ŸŽฏ Reinforcement Learning for Preference Optimization

Exploring RL techniques (RLHF, DPO) for preference learning to align LLMs with human values and make fine-tuning more accessible

๐Ÿ›ก๏ธ Deepfake Audio Detection Research

Researching robust neural audio fingerprinting for detecting synthetic music and deepfake audio using music foundation models

๐Ÿ” Advanced Information Retrieval System

Built enterprise-grade semantic search and retrieval system with vector databases for real-time similarity matching

๐Ÿค– LLM Fine-Tuning for Text Summarization

Fine-tuned Qwen2.5-0.5B using LoRA adapters and GRPO training with comprehensive evaluation framework

๐ŸŽต Audio Multimodality & AI Safety

Exploring neural audio fingerprinting and authentication systems to ensure trustworthy AI-generated content


๐Ÿ“ซ Let's Connect!


๐Ÿ† Certifications & Education

  • ๐ŸŽ“ MSc in Artificial Intelligence
  • ๐ŸŽ“ B.Tech in Computer Engineering (CGPA: 9.83/10)
  • โ˜๏ธ Microsoft Azure Fundamentals (AZ-900)
  • โ˜๏ธ Google Cloud Computing Foundation
  • ๐Ÿค– Prompt Engineering for Generative AI (LinkedIn Learning)
  • ๐Ÿ Python for Data Science (IBM)
  • ๐Ÿงฎ Introduction to Data Science (IBM)

๐ŸŽฏ Fun Facts

  • ๐Ÿ›ก๏ธ Passion: Building safe and trustworthy AI systems through deepfake detection and robust retrieval
  • ๐ŸŒ Mission: Making AI systems secure, reliable, and beneficial for society
  • ๐Ÿ“š Reading: Latest research papers on AI safety, deepfake detection, and neural audio processing

Connect with me:

twitter linkedin facebook instagram


Support My Work:

Buy me a coffee


โญ๏ธ From HARISHKUMAR1112001 with โค๏ธ

Popular repositories Loading

  1. restapiofuser restapiofuser Public

    This is the rest API for CRUD operation for create, delete, read and update user. user will have a name, username and password. you can't update password but just the name and username.You can tesโ€ฆ

    JavaScript 1

  2. Azure-OpenAI-Price-Calculation Azure-OpenAI-Price-Calculation Public

    This SDK provides an easy-to-use interface to fetch Azure OpenAI pricing data for specific models, deployment types, and locations. It filters relevant pricing details from the Azure and returns toโ€ฆ

    Python 1

  3. intelligent-content-evaluator intelligent-content-evaluator Public

    A production-ready Python framework for evaluating text content using multiple metrics including cosine similarity, BERTScore, and LLM-based evaluation with Azure OpenAI. Supports structured assessโ€ฆ

    Python 1

  4. cifar100-multimodal-embeddings cifar100-multimodal-embeddings Public

    Multimodal embedding pipeline for CIFAR-100: Skip-Gram text embeddings (86.9% MRR), CLIP-style image encoder (77.77% I2T Recall@1), and neuro-symbolic reasoning. Includes 5-phase BERT-guided genetiโ€ฆ

    Python 1

  5. HARISHKUMAR1112001.github.io HARISHKUMAR1112001.github.io Public

    HTML

  6. PythonPractical PythonPractical Public

    This is the basic python programe .

    Python

0