Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

AI/ML & software engineer with 4 years of experience building production AI inference systems, applied machine learning workflows, and scalable data pipelines. Skilled in Python, PyTorch, Hugging Face, and MLOps practices across Docker, CI/CD, AWS, and REST-based integrations. Experienced in performance evaluation and reliability engineering, including reducing real-time inference latency by 3–12× through system-level optimization. Holds a Master of Data Science (UWA) and is available immediately with full Australian working rights.…AI/ML & software engineer with 4 years of experience building production AI inference systems, applied machine learning workflows, and scalable data pipelines. Skilled in Python, PyTorch, Hugging Face, and MLOps practices across Docker, CI/CD, AWS, and REST-based integrations. Experienced in performance evaluation and reliability engineering, including reducing real-time inference latency by 3–12× through system-level optimization. Holds a Master of Data Science (UWA) and is available immediately with full Australian working rights.

Shilpa Sandhya

AI Engineer, Data Scientist, Software Engineer, +2





AI/ML & software engineer with 4 years of experience building production AI inference systems, applied machine learning workflows, and scalable data pipelines. Skilled in Python, PyTorch, Hugging Face, and MLOps practices across Docker, CI/CD, AWS, and REST-based integrations. Experienced in performance evaluation and reliability engineering, including reducing real-time inference latency by 3–12× through system-level optimization. Holds a Master of Data Science (UWA) and is available immediately with full Australian working rights.…AI/ML & software engineer with 4 years of experience building production AI inference systems, applied machine learning workflows, and scalable data pipelines. Skilled in Python, PyTorch, Hugging Face, and MLOps practices across Docker, CI/CD, AWS, and REST-based integrations. Experienced in performance evaluation and reliability engineering, including reducing real-time inference latency by 3–12× through system-level optimization. Holds a Master of Data Science (UWA) and is available immediately with full Australian working rights.

Available to hire

AI/ML & software engineer with 4 years of experience building production AI inference systems, applied machine learning workflows, and scalable data pipelines. Skilled in Python, PyTorch, Hugging Face, and MLOps practices across Docker, CI/CD, AWS, and REST-based integrations.

Experienced in performance evaluation and reliability engineering, including reducing real-time inference latency by 3–12× through system-level optimization. Holds a Master of Data Science (UWA) and is available immediately with full Australian working rights.

Skills

Work Experience

AI/ML Engineering Intern at Norwood Systems Ltd

July 1, 2025 - July 1, 2025

Migrated production speech-to-text inference from Python to C++23 using Whisper.cpp, achieving 3–12× lower inference latency for real-time applications. Built automated benchmarking workflows to evaluate latency, throughput, and runtime performance, enabling regression analysis and evidence-based optimization. Implemented production-ready API workflows for real-time audio processing and integrated inference components with downstream systems. Performed structured troubleshooting and root-cause analysis for integration, runtime, and performance issues to improve reliability in collaboration with senior engineers.

Senior Software Engineer at Tata Elxsi Private Limited

October 1, 2019 - May 1, 2023

Developed and maintained production-grade C/C++ software for the RDK-B Home Gateway platform used across millions of devices. Delivered maintainable fixes through structured debugging and root-cause analysis for complex production and integration issues across networked components. Led a 4-member team on an IPv6 stabilisation initiative, coordinating debugging, testing, and defect resolution to improve defect closure by 30%. Collaborated via technical discussions and peer review to deliver reliable end-to-end solutions, and contributed production fixes upstream to the open-source Dibbler DHCPv6 project through Git-based development with merged reviews.