The recruiter's phrase "language agnostic" is doing a lot of work, and reading between the lines, it's telling you something important: this interview is probably not about algorithmic gymnastics. It's about how you think with data. If you're torn between grinding LeetCode and polishing your Pandas and SQL, the honest answer is that your time is better spent on the latter, but not for the reason you might expect.
Let's be direct: most roles that mention "language agnostic" are signaling that they care about your problem-solving approach, not your ability to invert a binary tree from memory. That phrase typically appears in job descriptions for data-centric roles where the interviewer wants to see how you translate a messy business question into a clean, reproducible query or transformation. The person who posted this is right to guess it's not heavily DSA focused. In practice, that means you should expect questions that test your ability to manipulate data under time pressure, explain your logic out loud, and handle ambiguity. LeetCode-style problems often reward memorization and pattern recognition, but they don't tell an employer whether you can join three tables without dropping rows or reshape a dataset for a stakeholder who doesn't know what a pivot table is.
So what does this mean for you practically? Shift your prep toward scenario-based exercises. Open a dataset you've never seen before and practice writing SQL queries that answer specific business questions. Use Pandas to clean a messy CSV and then explain why you chose one method over another. Record yourself walking through your thought process. The interviewer wants to hear you reason through edge cases, handle nulls, and justify your choices. If they wanted a human compiler, they'd say so. They didn't. They said "language agnostic," which is code for "we want to see how you approach a problem, not whether you've memorized the solution."
One more thing: don't ignore the soft skills side. A language-agnostic interview often includes a live coding or whiteboard session where communication matters as much as correctness. If you can't explain why you're using a window function or why you'd rather do a merge in Pandas than a loop, you'll lose points even if the code runs. Practice talking about your data decisions as if you're presenting to a non-technical manager. That's the skill that separates someone who can write a query from someone who can drive a decision with it. So, skip the binary tree, pick up a messy dataset, and get comfortable being wrong out loud. That's the interview they're actually running.