📫 Contact: [email protected]
📞 Phone: +91 942-963-9215

I am passionate about Artificial Intelligence. Working with data makes me feel like a data detective, finding patterns and using them to help the world grow. Currently, I'm working on training and fine-tuning large language models like LLaMA, Falcon, Bloom, etc., on custom data. I believe in balancing research and implementation. Accuracy is important, but so is designing and developing scalable models. I am actively learning about Generative AI, Reinforcement Learning, MLOps, and AI System Architecture.

• 🔭 I’m currently working on training and fine-tuning large language models like LLama, Falcon, and Bloom.
• 🌱 I’m currently learning more about Generative AI, Reinforcement Learning, and AI System Architecture.
• 👯 I’m looking to collaborate on innovative AI projects that can make a significant impact.
• 🤔 I’m looking for help with advanced quantization techniques and model optimization.
• 💬 Ask me about AI model training, deployment, and NLP systems.
• 📫 How to reach me: [email protected]
• 😄 Pronouns: He/Him
• ⚡ Fun fact: I love combining my passion for AI with my detective-like curiosity to uncover hidden patterns in data.

• Training and Fine-Tuning: Pre-trained and fine-tuned large language models like LLama, Falcon, and Bloom on custom datasets for clients.
• Deployment: Deployed these models on client infrastructures, ensuring efficient inference and data privacy using quantization techniques to reduce memory footprints.

• Search Pipeline Update: Upgraded an existing keyword-based search pipeline to an embedding-based search pipeline using transformer models and Elasticsearch.
• System Development: Led a team to develop a document Q&A system from scratch, deployed it using Kubernetes.

• HR Systems: Developed and deployed systems to help HR filter candidates with good communication skills using Kubernetes.

• Project Development: Led a team to develop and deploy modules for a JD-CV comparison project.

• Chatbot Development: Developed and deployed chatbots using Bot Framework, Azure bot service, and rasa, including re-architecting systems for dynamic chat flow creation.
• Resume Parsing: Trained document layout and NER models to extract relevant data from resumes in a structured format.

• Pipeline Enhancement: Replaced an old OpenCV-based video processing pipeline with a deepstream-based pipeline.
• Model Training: Collected data, cleaned and preprocessed it, and trained models for object detection, classification, and tracking.

• End-to-End Ownership: Took ownership of the end-to-end speech-to-text process, transitioning from Kaldi to deepspeech for improved accuracy and efficiency.

• LinkedIn: Krupal Purohit
• GitHub: krupalkp97