Singapore, Singapore
https://sss2107.github.io/BuildWithAI/
I build intelligent solutions using GenAI and have worked on below problems: - Built an MCP-based AI integration connecting LLM to SAP S/4HANA, enabling natural language processing for automated maintenance task creation with secure OData v4 API workflows. - Designed and developed an advanced RAG chatbot, indexing over 10 million pages of all Airbus and Boeing aircraft manuals. Successfully implemented efficient query retrieval, processing, and caching mechanisms to achieve improvements in accuracy and latency. - Built model to predict aircraft recurring defects and recommend action for engineers - Built a multi-agent AMS chatbot connecting application/infrastructure logs with LLMs to autonomously diagnose and resolve production issues, significantly reducing manual intervention. - Extensive deployment experience of GenAI apps on in-house infra using Lambdas, EC2, Sagemaker pipelines, Dynamo DB. - Developed a confidential GenAI app for CEO meetings and ensured end-to-end secrecy through a zero-logging policy, preventing any response or log data from being transmitted to external LLM provider or cloud. - Contributed to Singapore Airlines GenAI Framework, defining implementation roadmap for AI adoption across the organization, and ensuring alignment with business objectives. - Mentored team of 5 data scientists and engineers. - Framework/Tools/Software: Pytorch, Tensorflow, AWS Sagemaker, Langchain, Llamaindex, OpenAI, Huggingface, Claude, Mistral, FastAI, AzureML Keras, Vertex AI, Docker, MLFlow, DVC, AWS Opensearch, FAISS, AWS Textract, PySpark, Django, Flask, Streamlit.
Best research paper award (OCPA, NUS, 2022). Capstone Project through National University of Singapore (Mentor: Prof. James Pang)- In this role, I experimented with free medical text such as doctor diagnoses and medical prescriptions to increase the accuracy of the existing Machine Learning models at Munich Re. I have solved ICD-10 code classification on free medical text using NLP techniques, Deep Learning, and Transformer models. Experimenting through different supervised and unsupervised techniques, the model gets an overall f1 score of 84% on unseen data including clinical text in multiple languages. I’m reproducing, testing, and enhancing algorithms from research papers to better the results.
• Developed pipelines for X0PA AI Room, a video interview platform using NLP and ML techniques. • Extracted features from various data sources to enable AI-based candidate scoring. • Collaborated with team to prototype and implement scoring metrics based on text transcripts and sentiments.
In my role as an AI Researcher at the National University of Singapore, I conducted research on Incentive Mechanism for Federated Learning and DeepChain. By applying machine learning algorithms with homomorphic encryption, I ensured data privacy during model training. Collaborating with Prof. James Pang and Prof. Khor Ping, we developed innovative solutions in the field of AI research.
• Defined ICT strategies and managed technology solutions for business challenges in FS and TMT industries. • Led teams in executing technology implementations such as ERP, BI, digitization, analytics, and cloud. • Contributed to white papers on telecom sector disruption and cloud adoption in the Gulf region.