Behind the Mic with Andrea Er Software Engineer at the Government Technology Agency of Singapore :

Andrea Er is a Software Engineer at the Government Technology Agency of Singapore (GovTech), formerly deployed as an AI specialist to the JTC Corporation’s Smart District Division. She possesses extensive experience in managing the complete lifecycle of diverse AI projects. Her portfolio showcases a wide range of AI solutions, ranging from enhancing facility management to integrating AI capabilities into the Open Digital Platform (ODP) for the Punggol Digital District.

We had the privilege of speaking with Andrea Er about her journey, experiences, and perspectives in the world of AI, here are her thoughtful responses.

Q1. What inspired you to pursue a career in AI, and what key milestones have shaped your journey in this field?

My journey into AI began many years ago through hands-on work in robotics research, which gradually evolved into a broader exploration of intelligent systems. A pivotal chapter was during my deployment to JTC Corporation’s Smart District Division under GovTech, where I worked on enabling autonomous robots to navigate complex indoor environments and visualise themselves within digital twins. This required deep integration with building infrastructure such as lifts, turnstiles, and automated doors, particularly across multiple floors.

I started by using the lift as a fixed reference point, applying regression models to align the robot’s maps with real-world coordinates. While this approach was effective, the dynamic nature of the environment led me to explore feature-based mapping to identify consistent landmarks in the surroundings and maintain better spatial awareness even as conditions changed.

In addition, I also designed a robot queue management system to allow multiple robots to navigate safely within a shared space, including route planning and conflict resolution. This deepened my interest in learning-based approaches and real-time decision-making. Motivated to go further, I began studying machine learning and computer vision in my own time, drawn by the potential for AI to reason under uncertainty and learn from data directly.

The turning point was when my division decided to form an AI-focused team. This gave me an opportunity to apply what I had learned to real-world projects, shifting from solving isolated technical problems to developing AI solutions that supported smart urban systems and improved day-to-day building operations.

What continues to inspire me is AI’s ability to help us understand complex environments and make better decisions. I’m driven by the belief that AI, when thoughtfully applied, can help us navigate complexity and build environments that work intuitively for everyone.

Q2. What are some of the most significant challenges you've faced when designing or deploying AI models, and how did you overcome them?

One of the key challenges in designing and deploying AI models was dealing with inconsistent and noisy data from a wide range of IoT devices and legacy systems. This variability made it difficult to maintain reliable model performance. To address this, I developed standardised processes to clean and normalize incoming data, to ensure a consistent baseline for the AI models to work effectively.

Another significant challenge was data drift, where changes in input data over time may degrade model accuracy. To tackle this, I designed and implemented an end-to-end drifting monitoring system that continuously tracks key metrics such as accuracy, drift, and latency. This system alerts developers when performance dips below a preset threshold thus enabling timely intervention. Complementing this, I put in place automated retraining pipelines that update models the moment degradation is detected, helping to keep AI solutions reliable under real-world conditions.

To strengthen model resilience, I also developed synthetic data generators that simulate edge cases during development. This allowed our team to prepare AI models for a wider variety of scenarios and unexpected conditions. Establishing uniform data formats and flexible frameworks further enabled efficient deployment and customisation across multiple projects.

These efforts have been successfully deployed to production, allowing the AI team to maintain performance, reduce downtime, and ultimately deliver AI solutions that provide tangible value in smart urban environments.

Q3. What strategies have you found most effective for ensuring AI solutions deliver meaningful, ethical, and user-centric outcomes?

I focused on designing solutions that are both strategically aligned and operationally grounded. The goal is not to simply deploy advanced technology, but to ensure it delivers tangible impact, be it by reducing energy consumption, streamlining operations, or enhancing user experiences.

An example is our work on optimising energy use in building systems. We developed models to improve the performance of air-conditioning and mechanical ventilation (ACMV) systems. Leveraging real-time sensor data and optimisation algorithms, these models dynamically recommend cooling tower setpoints based on occupancy patterns and ambient environmental conditions. These models balance thermal comfort with overall plant efficiency, accounting for system-wide trade-offs rather than isolated components.

Beyond technical design, we invested in frameworks that support consistent and transparent deployment. For instance, I authored the first version of the Open Digital Platform (ODP) Robotics Playbook, a guide for external robot vendors to onboard and integrate their robots with the ODP. This playbook defines clear processes on how robotic systems can interact with our infrastructure, allowing us to manage them both reliably and at scale. It also promotes alignment among stakeholders through a shared approach, which is essential for sustainable adoption.

For me, ethical AI is not an afterthought, it is embedded in every part of the process. From setting design goals to evaluating outcomes. Values such as fairness, accountability, and transparency should guide AI development at every stage. This approach respects the people impacted by these systems and ensures that AI delivers real, lasting benefits to the communities it serves.

Q4. How do you collaborate with cross-functional teams, such as engineering, design, and marketing, to ensure seamless customer experiences?

Effective collaboration across diverse teams is essential to delivering integrated and seamless experiences, especially within the complexity of a smart district environment. In my role, I prioritise working closely with stakeholders from engineering, design, and other functions to ensure technical solutions address user needs while supporting wider organisational objectives.

For example, as the project manager for the cooling tower optimisation application, I coordinated the efforts of both backend engineers and digital twin specialists. My role involved aligning both teams toward a common goal and establishing standardised communication protocols. These protocols enable smooth data exchange between AI and machine learning systems, IoT devices, and other platforms, even when they were built by different teams.

Beyond technical collaboration, I have worked extensively on integrating cyber-physical systems such as robotics and lift management. This required a close partnership with teams managing physical infrastructure to ensure that control systems and physical components operate in harmony.

Having experience in both backend development and frontend interfaces gives me a comprehensive understanding of the entire system. This perspective allows me to bridge gaps between different teams effectively and maintain clear communication. By fostering mutual understanding and shared purpose, I help drive projects forward and deliver cohesive solutions that improve the everyday experience for everyone in the smart districts we develop.

Q5. How do you stay ahead of rapidly evolving trends in AI, generative technologies, and digital transformation to inform your strategy?

Staying ahead in AI requires more than just keeping up with the latest trends. It is a continuous process of learning, experimenting, and sharing. I often read research papers, attend technology conferences, and exchange ideas with peers working in similar spaces. These help me understand where the field is heading and how emerging concepts can be applied meaningfully in the work we do.

However, staying informed is only a part of the process. I make it a point to try things out through hands-on experimentation. For example, I have developed and deployed Vision-Language Model to analyse CCTV feeds and provide location-specific insights to enable real-time decision-making. It is specifically tuned to detect safety-related incidents, such as unattended objects and signs of fire. Its ability to classify events accurately reduces false alarms and supports more efficient use of manpower.

I also implemented a computer vision system to monitor car park gantries for stalled vehicles. The tool alerts facility managers in real time, helping them intervene quickly and manage congestion effectively. These experiments help us ground new technologies in actual use cases that matter to operations.

Sharing what I learn with my team is equally important. I regularly present key insights and working prototypes so we can learn together and incorporate new knowledge into our strategy. This ensures our work remains relevant, grounded, and responsive to change.

In a fast-moving field, it is easy to tunnel vision on the latest and greatest. However, I try to stay focused on what is useful. My priority is to understand how these technologies can improve public services and contribute to systems that are thoughtful, efficient, and built to last.

Conclusion

Andrea Er’s journey is a powerful example of how curiosity, hands-on experience, and a deep sense of purpose can drive meaningful innovation in AI. From enabling smart robotics in urban infrastructure to pioneering responsible AI deployment at scale, Andrea exemplifies what it means to build technology that is both cutting-edge and grounded in human impact.
Her thoughtful integration of ethics, cross-functional collaboration, and practical experimentation reflects a future-forward mindset that prioritises sustainability, transparency, and value for all stakeholders. As AI continues to evolve, professionals like Andrea remind us that true progress lies not just in what we build, but in how and why we build it.

“Stay tuned for more such insights in our upcoming Behind the Mic editions — where we bring you closer to the minds shaping the future of technology and experience.”