On January 7, 2019, I started my first full-time job. Eight years later, I would like to begin the next full-time chapter around the same date. I like the symmetry. Maybe it is arbitrary, but I think it is okay for a date to be personally meaningful without making it a rigid deadline or a grand prediction.

It is not a fixed promise to myself. It is more like a constraint that helps me hold this period with a little more intention.

Between now and then, I have a very interesting stretch of time. This fall semester has come together in a way that gives me both structure and room to explore. I am teaching / supporting work around Foundation Model Engineering, and I am also spending time around agentic systems and human-AI systems. I am learning a lot through the course material, through discussions with students and peers, and through trying to explain ideas clearly enough that they can be useful to someone else.

At the same time, I have space to think.

As a student, I sometimes feel like I have a full day, perhaps twelve or more hours, to think about different aspects of AI, systems, research, work, people, technology, and the directions that are opening up. Some days that means reading deeply. Some days it means building prototypes or doing end-to-end projects. Some days it means writing notes, doing thought experiments, talking to people, revisiting old ideas, or getting stuck on questions that do not yet have a clean answer.

At times, it can feel like wandering. There are too many papers, too many models, too many product directions, too many possible projects, too many interpretations of what “the future of AI” should look like. But I also think that wandering is part of the process.

I do not want to follow only a predefined recipe. I do not want to move from one checklist item to another simply because that is what a standard career roadmap says to do. I want to give myself permission to enter uncertain spaces, struggle with difficult questions, build things that may not work, test intuitions against reality, and learn from the outcomes.

The work I keep returning to sits somewhere around AI/ML, data science, research engineering, agentic systems, evaluation, reliability, safety, and human-AI interaction. I am interested in the model layer, but I am equally interested in everything around it: memory, tools, context, evaluation, observability, failure modes, coordination, incentives, workflows, and the ways humans and AI systems interact in real environments.

I am especially interested in systems that have to survive outside a demo. Systems that need to be useful, reliable, inspectable, adaptive, and grounded in real constraints. Systems where it is not enough for a model to produce an impressive answer once. It has to interact with tools, data, people, uncertainty, feedback, changing environments, and sometimes conflicting objectives.

The last year has given me a lot to think about. Graduate school, technical work, experimentation, coursework, the internship experience, conversations with different people, and looking back at my previous full-time work have all changed how I think about what I want to do next. I have had encouraging leads and good conversations through my internship, and I am grateful for them. But I also want to use this period to explore further before committing to the next role.

I want to find problems that are genuinely worth spending years on, not simply a title or an available position. I want to find a setting where my past experience in applied AI/ML, data science, product work, research prototyping, and system-building can interact with the things I am now learning more deeply about foundation models, agents, evaluation, and responsible AI.

I have been thinking about this whole transition as something like a complex adaptive system.

There are many components in it. There is my prior experience. There are the skills I have built and the ones I still need to build. There is time, work authorization, energy, health, finances, responsibilities, available opportunities, timing, people, and luck. There are also my own assumptions, preferences, ambitions, blind spots, and biases.

Then there are the interactions.

A conversation might change how I see a problem. A small project may turn into a serious research direction. A rejection may reveal that I need to improve something. An introduction may lead to an unexpected collaboration. A paper may create a new question. A failed prototype may teach more than a successful one. A teaching conversation may force me to clarify an intuition I had never properly articulated.

Some of these interactions may create positive feedback. Some may create negative feedback. Some may lead nowhere. Some may open a direction that I cannot currently see.

That is why I do not see this as a linear roadmap. I see it as a continual optimization process under real constraints, with partial information, uncertainty, feedback loops, and changing objectives.

I am one agent in a much larger ecosystem. Other people have their own objectives, constraints, incentives, interests, skills, and paths. There are collaborations, conflicts, coordination problems, opportunities, institutions, market changes, technological shifts, and broader forces that influence what becomes possible. The current pace of AI development makes this especially visible. Some roles are changing quickly. Some forms of work are becoming less interesting or more automated. At the same time, entirely new problem spaces, tools, research questions, and opportunities are emerging.

I do not think anyone has a complete map of where this is going.

For me, the important thing is not to become passive inside that uncertainty. It is to stay curious, keep building, keep learning, reflect honestly, and adjust when the evidence changes.

And even if one project fails, one application does not work out, one lead goes cold, or one plan changes, I do not think the system collapses. There are enough stabilizing components: experience, skills, curiosity, discipline, relationships, learning capacity, and the ability to continue. The goal is to remain stable enough to keep moving, while staying open enough for new information to change the direction.

So this is not a polished announcement or a performative roadmap.

It is more like placing a signal into a system and seeing what it interacts with.

Over the next few months, I plan to build, experiment, read, write, teach, learn, share parts of the process, and keep looking for the questions and problem spaces that feel both difficult and meaningful.

If any of this resonates with you, if you are working on related problems, if you are thinking about agentic systems, foundation models, evaluation, AI safety, human-AI interaction, applied AI, or research engineering, I would be glad to connect.

If you think there is a problem, team, role, project, person, idea, or conversation that may be relevant, I would genuinely appreciate hearing from you. Advice, critique, collaboration, introductions, opportunities, disagreement, and perspective are all useful signals.

I will share what I learn and build along the way.

The goal is not to follow a perfect plan. It is to engage seriously with the system, learn from the interactions, and see what emerges by the time I reach the next starting line.