◦ Developed an ML training framework with embeddings.◦ Developed LLM APIs with LangChain and agentic interactions with LangGraph.◦ Managed search and retrieval tasks with embeddings using Vectorstores.◦ Designed a framework for validating LLM responses and performing LLM APIs' load testing.◦ Explored messaging queues for scaling LLM API requests.◦ Technologies: Python, Langchain, Langgraph, VectorStores, Retrievers, Pandas, GCP, Fast API, Scikit, Github Workflows, NATS… Show more ◦ Developed an ML training framework with embeddings.◦ Developed LLM APIs with LangChain and agentic interactions with LangGraph.◦ Managed search and retrieval tasks with embeddings using Vectorstores.◦ Designed a framework for validating LLM responses and performing LLM APIs' load testing.◦ Explored messaging queues for scaling LLM API requests.◦ Technologies: Python, Langchain, Langgraph, VectorStores, Retrievers, Pandas, GCP, Fast API, Scikit, Github Workflows, NATS messaging queues. Show less

◦ Built asynchronous pipelines for web scraping, data preprocessing using ML and NLP techniques.◦ Enabled Retrieval-Augmented Generation (RAG) with vector databases and GenAI frameworks for optimized search.◦ Engineered extensible APIs to support future LLMs and diverse prompt formats.◦ Developed a modular prompt-management framework for precise model guidance.◦ Fine-tuned open-source LLMs on multi-GPU setups using Accelerate, DeepSpeed, and FSDP frameworks.◦ Collaborated… Show more ◦ Built asynchronous pipelines for web scraping, data preprocessing using ML and NLP techniques.◦ Enabled Retrieval-Augmented Generation (RAG) with vector databases and GenAI frameworks for optimized search.◦ Engineered extensible APIs to support future LLMs and diverse prompt formats.◦ Developed a modular prompt-management framework for precise model guidance.◦ Fine-tuned open-source LLMs on multi-GPU setups using Accelerate, DeepSpeed, and FSDP frameworks.◦ Collaborated with business teams to design and evaluate innovative product features.Technologies: Python, Pytorch, Pandas, AWS, Fast API, AsyncIO, Jinja, Redis, REST, Git, Docker,Qdrant Vector DB, Scikit, nltk, LangChain, Streamlit Show less

Smart Data Connector◦ Contributed to developing Smart Data Connector, an Enterprise SaaS for building and maintaining Knowledge Graphs (KG).◦ Developed GPT-3-based triple verbalization, reducing time by ~50%.◦ Integrated KG based QA and KG Entity Disambiguation with BART classifiers to ensure factual correctness in LLM responses.OpenGPT-X◦ Contributed to OpenGPT-X, developing a data framework for multilingual LLM pre-training.◦ Designed and developed a modular NLP… Show more Smart Data Connector◦ Contributed to developing Smart Data Connector, an Enterprise SaaS for building and maintaining Knowledge Graphs (KG).◦ Developed GPT-3-based triple verbalization, reducing time by ~50%.◦ Integrated KG based QA and KG Entity Disambiguation with BART classifiers to ensure factual correctness in LLM responses.OpenGPT-X◦ Contributed to OpenGPT-X, developing a data framework for multilingual LLM pre-training.◦ Designed and developed a modular NLP framework (AQuA) for scalable document quality management.◦ Authored reports on data quality with semantic and syntactic perspectives.Technologies: Python, Pandas, Sci-Kit, PyTorch, Knowledge Graphs, GIT, Nlt-K, Docker, PySpark, KubeFlow, AWS, Numpy, Jupyter. Show less

Interdisciplinary work conducted in collaboration and supervision of the Psybernetics Lab research group. The group focuses on HCI, Machine Learning and Cognitive Psychology and develops AI technologies that enhance cognitive capabilities such as learning and reasoning. My work dealt with researching and developing ideas to analyze and estimate the cognitive states of a Reader using Machine and Deep Learning techniques. These insights were then used by a Language model to recommend… Show more Interdisciplinary work conducted in collaboration and supervision of the Psybernetics Lab research group. The group focuses on HCI, Machine Learning and Cognitive Psychology and develops AI technologies that enhance cognitive capabilities such as learning and reasoning. My work dealt with researching and developing ideas to analyze and estimate the cognitive states of a Reader using Machine and Deep Learning techniques. These insights were then used by a Language model to recommend personalized content to the reader as per their level of comprehensibility. Technologies: Python, PyTorch, Sci-kit, Pandas, nltk Show less

◦ KYTT (Know Your Text): Built RDF Knowledge Graphs from text using parsers, First Order Logic, and Entity Disambiguation.◦ ReNewRS: Developed modular web crawlers for article management for the research project that focussed on Responsible AI to mitigate polarization effects in news recommender systems.Technologies: Python, RDF, Docker, nltk, Scrapy, Spacy, Fast API

◦ Conducted research on preventive QA, creating image pipelines to detect early production defects.◦ Leverage landmark detection, image segmentation, and unsupervised deep learning (Autoencoders, Clustering) for anomaly detection.◦ Authored report on unsupervised defect detection techniques in images of apparels in the wild.Technologies: Python, TensorFlow and PyTorch, Pandas, Open-CV, Flask

Nirathisayam is a boutique business in the custom framing and gifting space.- Develop a web-based order management system with CRM features.Technologies: PHP, JavaScript, SQL, JQuery.

◦ Built APIs to aggregate data from multiple sources within Sprinklr, improving accessibility.◦ Developed middleware for third-party integrations, expanding platform capabilities.◦ Prototyped a data enrichment system, boosting Marketing's lead collection efficiency by 4x.◦ Prototyped a tool automating custom data delivery, reducing response time and improving client satisfaction.◦ Led migration of legacy Perl scripts to class models, enhancing maintainability and… Show more ◦ Built APIs to aggregate data from multiple sources within Sprinklr, improving accessibility.◦ Developed middleware for third-party integrations, expanding platform capabilities.◦ Prototyped a data enrichment system, boosting Marketing's lead collection efficiency by 4x.◦ Prototyped a tool automating custom data delivery, reducing response time and improving client satisfaction.◦ Led migration of legacy Perl scripts to class models, enhancing maintainability and scalability.Technologies: Java, Spring MVC, Pandas, jsoup, MongoDB, ElasticSearch, SQL Show less

- Develop a crawler based framework for providing insights about Job postings based on location, skills, and other metrics. The data was processed, and structured in the Hadoop environment.- Technologies used - Java, HDFS, HiveQL, JavaScript.