Data Scientist @AiLogic | Ex-MTS @Zoho | NLP | LLM | ML | GATE’23 CSE Top 10%
As a Senior Data Scientist with a robust foundation in Artificial Intelligence, Machine Learning, Natural Language Processing, and Generative AI models, I have dedicated my career to developing and optimizing complex systems. At Zoho, I specialized in Neural Machine Translation and Transliteration, enhancing models crucial for internal product teams to effectively meet user needs.
In my current role at Ailogic, I remain informed about the latest advancements in AI by regularly engaging with contemporary research and experimenting with emerging models to benchmark their applicability. My hands-on approach includes building machine translation models from the ground up, encompassing tokenizer training and developing PyTorch-based architectures. I employ inference optimization techniques such as KV cache, multiprocessing, and distillation, alongside leveraging Ray along with Docker and Kubernetes to ensure scalability, containerize and orchestrate models, ensuring consistent and efficient operations across various environments.
My expertise extends to fine-tuning models using Supervised Fine-Tuning (SFT) and Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA and QLoRA, facilitating effective model quantization. I have also architected AI servers, conducting comprehensive benchmarks of top GPUs—including DDR and SXM variants—to efficiently scale infrastructure.
Additionally, I have implemented Retrieval-Augmented Generation (RAG) techniques to develop specialized internal chatbots, enhancing information retrieval within the organization. Furthermore, I have designed Agentic AI systems—multi-utility bots capable of accessing internal databases, reading and writing files, and performing web searches as needed. These systems utilize LangChain-based frameworks to manage memory caches, ensuring efficient and contextually relevant interactions.
My commitment to the AI community is reflected in my open-source contributions, notably addressing critical bugs in projects like ONNX Runtime and resolving logic issues in initiatives such as ACES and PYTHAINLP, leading to merged pull requests and new patch releases. These efforts have been recognized with GitHub's Pull Shark badge, awarded for successfully merging multiple pull requests.
Passionate about leveraging data to drive strategic decisions, I am dedicated to advancing AI technologies and their practical applications.
Work Terms
Could work for upto 20 hours a week, with starting a minimum of $20/hour could be negotiable based on the project, Preferred mode of language would be English, Preferred mode for reaching out would be my official Mail - varunkatiyar819@gmail.com