The question I keep hearing from Indonesian engineers considering a master's degree is some version of "should I study AI." The honest answer is: probably not, unless your goal is to do research. The more useful question is which specialities become more valuable precisely because AI is commoditizing the code-writing part. I have spent time thinking about this because I went through the same decision process, and the conclusion I landed on surprised me.
Most engineers default to what feels safe. A master's in computer science, a generalist degree, maybe machine learning because the job market looks hot. But the engineers I know who are genuinely thriving are not the ones who learned to train models. They are the ones who learned to build the systems around the models, or who moved into the problem spaces where code alone was never the bottleneck.
What is AI actually replacing?
AI is getting very good at translating intent into code. Given a clear specification, a capable model can produce working implementations across most standard patterns. CRUD endpoints, data transformations, UI components, test scaffolding. The pattern-matching layer of software engineering is being compressed.
What AI is not replacing is the layer above and below that. Above: understanding what to build, why, and for whom. Below: making systems that actually work at scale, under real constraints, with real failure modes. The specification layer and the infrastructure layer are both getting more valuable, not less.
Why product engineering and human-computer interaction?
If AI handles the implementation, the engineer who understands the user problem becomes more valuable, not less. Product thinking, UX engineering, and the ability to design systems around human behavior are becoming the differentiator between an engineer who ships features and an engineer who ships outcomes.
The demand here is less about job titles and more about leverage. Product engineers, design engineers, and UX-focused full-stack roles are growing because companies are realizing that the bottleneck is not writing code. It is knowing what to build. Engineers who can bridge the gap between user research and production code are rare, and they tend to end up in senior roles faster because they make better decisions about what not to build.
What should you study?
MSc Human-Computer Interaction or Design Engineering. If your goal is to own the user-facing layer and make product decisions, this is more useful than another CS degree. The best programs combine design methods, user research, prototyping, and enough technical depth that you remain credible as an engineer.
Where do LPDP students go for this?
- University College London (UCL) (UK). The MSc in Human-Computer Interaction is one of the best in the UK. London's product and design ecosystem is a genuine advantage. One year.
- University of Glasgow (UK) and University of Bristol (UK). Both have well-ranked computing and interaction design programs. Lower cost of living than London and both appear on the LPDP eligible list.
- University of Melbourne (Australia). Strong HCI offerings within the computing and information systems school. 1.5 to 2 years.
- University of Washington (US). Strong in HCI, with Seattle's tech ecosystem providing real product and design industry exposure. Public university with relatively manageable costs.
- Carnegie Mellon University (US). The gold standard for HCI. The Human-Computer Interaction Institute is unmatched in depth and reputation. It is private and significantly more expensive. LPDP can cover it, but the budget impact is real. Worth it if you get in and the program aligns tightly with your goals.
Why data engineering and ML infrastructure?
Not building models. Building the pipelines, feature stores, monitoring, and deployment infrastructure that make models useful in production. Most ML projects fail not because the model is bad but because the data pipeline is unreliable, the feature engineering is inconsistent, or the serving layer cannot handle the latency requirements. This is an engineering problem, not a data science problem.
Demand is being driven by the AI boom itself. Every company deploying models needs data engineers and ML platform engineers. The gap between "we have a model in a notebook" and "we have a model in production" is almost entirely an engineering gap. Data engineering roles are among the fastest-growing in the industry, and they pay at or above general software engineering rates because the supply is thin.
What should you study?
MSc Data Science or Data Engineering. The programs worth doing are the ones that include substantial engineering content: data pipelines, distributed computing, systems for ML, not just statistics and Jupyter notebooks. If the curriculum is mostly R and hypothesis testing, it is a statistics degree, not a data engineering degree.
Where do LPDP students go for this?
- Imperial College London (UK). The strongest option for data science with an engineering orientation. Rigorous program and strong industry connections in London. One year.
- University of Edinburgh (UK). Data science programs benefit from the school's deep AI and informatics research. Strong balance of engineering and analytical content.
- University of Manchester (UK). Solid data science programs with lower cost of living than London. Strong engineering faculty.
- University of Melbourne (Australia). Well-balanced data science program within a strong computing school. 1.5 to 2 years.
- Australian National University (ANU) (Australia). Research-intensive orientation in computing and data. Canberra is affordable relative to Sydney and Melbourne.
- Georgia Tech (US). The analytics and computing programs have strong ML infrastructure and data systems tracks. Public university tuition.
- UIUC (US). One of the best programs in the country for data infrastructure and large-scale systems. Public university with reasonable costs.
Why product management?
This is the path for engineers who realize the highest-leverage work they do is not writing code. It is deciding what gets built, why, and in what order. Software engineers who move into product management carry an advantage most traditional PMs do not have: they can evaluate technical feasibility in real time, they know when an engineer is sandbagging an estimate, and they can prototype before committing a team to a quarter of work.
The demand for technical product managers is growing faster than the supply. Companies building AI products especially need PMs who understand the constraints of models, the cost structure of inference, and the difference between a demo and a product. An engineer who can hold both the user problem and the technical architecture in their head at the same time is exactly the profile these roles are designed for. Indonesia's own startup ecosystem, from Gojek to Tokopedia to the next wave, is producing PM roles that did not exist five years ago, and they strongly prefer candidates with engineering backgrounds.
What should you study?
MBA with a technology or product focus, or MSc in Technology Management / Innovation. A traditional MBA gives breadth in strategy, finance, and operations. The ones worth doing for this path are programs with strong tech electives, startup ecosystems, and product management coursework. If a full MBA feels too broad, some universities offer specialised MSc programs in technology management or digital innovation that are shorter and more focused.
Where do LPDP students go for this?
- Imperial College London (UK). The MSc in Management or the MBA has a strong technology and innovation orientation. London's startup and tech ecosystem is directly relevant. One year for the MSc.
- University of Edinburgh (UK). The Business School offers MSc programs in Management and Innovation that combine well with a technical background.
- University of Manchester (UK). The Alliance Manchester Business School has solid technology management programs. Lower cost of living than London.
- University of Melbourne (Australia). The Melbourne Business School MBA and the Master of Management have strong reputations. 1.5 to 2 years.
- UNSW (Australia). The AGSM MBA is well-regarded in the Asia-Pacific region with strong tech industry connections in Sydney.
- Georgia Tech (US). The MBA program has a technology focus that is hard to match at other business schools. It is a public university, so tuition is significantly lower than peer MBA programs. The combination of Georgia Tech's engineering reputation and a business degree is a strong signal.
- Carnegie Mellon University (US). The Tepper School MBA with a technology track, or the MS in Product Management (one of the few dedicated PM master's programs in the world). Private university costs apply.
Why computational science and domain-specific engineering?
Biotech, climate, fintech, robotics. These are fields where the domain knowledge is the moat. An engineer who understands both the code and the science can do things that a generalist with an AI copilot simply cannot. The degree serves double duty: it builds the domain expertise and signals seriousness to employers in that field.
The job demand is vertical-specific but deep. Fintech companies need engineers who understand financial instruments, not just APIs. Biotech companies need engineers who understand molecular data, not just data pipelines. Climate tech needs people who understand atmospheric modelling, not just model training. These roles are growing fast in their respective industries and they pay a premium because the intersection of engineering skill and domain knowledge is genuinely rare.
What should you study?
MSc in a domain-specific field like Computational Biology, Financial Engineering, Financial Technology, Robotics, or Climate Science. These degrees are most valuable when combined with prior software engineering experience because you bring the building skill and add the domain knowledge on top.
Where do LPDP students go for this?
- Imperial College London (UK). Strong programs in computational biology, financial technology, and environmental engineering. London's financial sector and biotech corridor are directly relevant.
- University of Edinburgh (UK). Research-intensive programs in computational science, bioinformatics, and robotics. One year.
- University of Melbourne (Australia) and UNSW (Australia). Both have strong applied science programs. Melbourne for biomedical and financial applications, UNSW for renewable energy and engineering applications.
- Georgia Tech (US). Excellent interdisciplinary programs in computational science, robotics, and bioengineering. Public university costs.
- Carnegie Mellon University (US). The Robotics Institute is world-leading. Financial engineering program is also strong. Private university budget constraints apply.
Why systems and infrastructure engineering?
Distributed systems, cloud architecture, reliability engineering. AI can write a Lambda function. It cannot design the failure modes of a distributed queue, reason about consistency guarantees under partition, or decide when to shard a database versus when to rethink the data model. This layer requires understanding physics, economics, and human operations simultaneously. It is not going away.
The demand is structural. Every company running production workloads needs people who can reason about uptime, scaling, and failure recovery. Cloud providers are growing, not shrinking. Platform engineering and site reliability engineering roles are consistently among the hardest to fill because the skill set takes years of real production exposure to develop. A master's degree accelerates that foundation.
What should you study?
MSc Computer Science with a systems or infrastructure focus. Look for programs that emphasize distributed systems, operating systems, networking, and cloud computing. The ones worth doing are programs where you are building and breaking systems, not just studying algorithms in isolation.
Where do LPDP students go for this?
- University of Edinburgh (UK). The informatics school is one of the best in Europe. Strong distributed systems and networking research. Edinburgh has a well-established Indonesian student community and one-year programs keep costs manageable.
- University College London (UCL) (UK). Strong systems track within the CS department. London is expensive but the breadth of the program and industry proximity are hard to match.
- University of Melbourne (Australia). Consistently the highest-ranked Australian university for computing. The Master of Computer Science has strong systems electives and Melbourne has a large Indonesian student community. Programs run 1.5 to 2 years.
- Georgia Institute of Technology (Georgia Tech) (US). One of the most realistic top US options for LPDP students. Excellent computing programs, lower tuition than most peers because it is a public university, and strong systems and networking research.
- University of Illinois Urbana-Champaign (UIUC) (US). One of the best CS programs in the country, particularly in systems and data infrastructure. Public university with a significant international student body.
Why security and privacy engineering?
The attack surface is growing faster than the defense tooling. AI introduces new categories of vulnerability: prompt injection, model poisoning, data leakage through embeddings. Security was already understaffed. It is about to be critically understaffed.
The job market reflects this. Cyber security roles have had near-zero unemployment globally for over a decade. Indonesia's own digital transformation, from GoPay to government systems, is creating domestic demand that did not exist five years ago. An engineer who understands both systems and security has one of the most defensible positions in the industry. This is also a field where the credential genuinely matters because employers in government, finance, and critical infrastructure filter for it.
What should you study?
MSc Cyber Security or Information Security. The best programs combine technical depth (cryptography, network security, penetration testing) with systems-level thinking (secure architecture, threat modelling, incident response). Avoid programs that are mostly policy and governance unless you want a non-technical career path.
Where do LPDP students go for this?
- University of Edinburgh (UK). Strong cyber security program with deep ties to the informatics school. One-year program.
- Imperial College London (UK). The computing security track is rigorous and benefits from London's financial sector, which is one of the largest employers of security engineers in Europe.
- University of New South Wales (UNSW) (Australia). Particularly strong in cyber security. The engineering faculty is well-regarded and Sydney has a growing security industry.
- University of Melbourne (Australia). Solid cyber security offerings within the computing school. Benefits from Melbourne's large financial services sector.
- Georgia Tech (US). The Information Security program is one of the best in the US. The research output in systems security is consistently top-tier and it remains affordable as a public university.
How do you choose between these?
The decision framework I would use:
- Start from the speciality, not the university. Decide which of the six areas above matches where you want to build your career. Then find the programs that are strongest in that area. Prestige without program fit is a waste of two years.
- Weight the program structure against your experience level. If you have less than three years of experience, a more structured taught program (common in the UK and Australia) is better than a research-heavy one. If you have five or more years, a program with a research component or thesis option lets you go deeper.
- Factor in the ecosystem, not just the classroom. An HCI degree in London or Seattle puts you near product-heavy companies. A systems degree near a cloud provider hub matters. The university is part of the equation. The city and the industry around it are the other part.
- Be honest about the LPDP budget. UK one-year programs leave more budget headroom. US two-year programs at private universities will stretch the funding. Australian programs sit in between. This is a real constraint that should inform your shortlist, not an afterthought.
What would I avoid?
A generic MSc in Computer Science with no specialisation track. The degree itself is fine, but the opportunity cost is high. Two years and full LPDP funding should buy you a specific, defensible expertise. "I have a master's in CS" does not differentiate you. "I spent two years building distributed systems at Edinburgh" or "I studied human-computer interaction at UCL" tells a story that maps to a career.
I would also avoid chasing an AI or machine learning master's purely because it sounds relevant to the moment. Unless your goal is ML research or you want to work specifically on model development, the applied engineering roles around AI do not require an ML degree. They require systems thinking, data engineering skill, and product judgment. Those are better served by the specialities above.
The AI era does not make software engineers less valuable. It makes generalist code-writers less valuable. The response is not to learn AI. It is to become the kind of engineer that AI makes more productive, not the kind it replaces.