Eating Ambiguity

The Problem: Integrating high-stakes, managed-access data into the Human Cell Atlas after a total team turnover (Director, PM, and Engineers) left the project with zero institutional memory and fractured stakeholder trust.
The Solution: Decoupling immediate compliance needs from technical debt, while establishing new transparency rituals to bridge the gap between engineering and legal stakeholders.
The Takeaway: When you’re dealing with ambiguity, your most powerful tools aren’t just technical — they are communication frameworks that re-establish trust and clarity.

Most people view a massive, undefined problem as chaos and freeze.
I see a complex, high-stakes project in need of decisive operational leadership.

When I was tasked with integrating managed-access data into the Human Cell Atlas at the Broad Institute, I wasn’t just inheriting a “tangle” of legal compliance and aging infrastructure — I was walking into the aftermath of a reorg. The previous Director, Project Manager, and Lead Engineers had all moved on, taking the institutional knowledge with them. Dealing with ambiguity at this scale was like catering a high-stakes party where the guest list was a mystery, the staff had disappeared, and the kitchen was in the weeds.

To navigate this without losing momentum, my first move was de-risking through decoupling. The initial mandate was a massive “migrate the system and make it compliant” combo. I knew that trying to execute both simultaneously with minimal staffing was a going to lead to gridlock and burnout. Instead, I separated the immediate business value — securing sensitive data access — from the long-term technical debt of a full system migration. By prioritizing the compliance framework first, we unblocked the researchers immediately and protected that high priority deliverable from the weight of the larger, more ambiguous project.

Because I was essentially starting from zero, I used AI as a force multiplier to accelerate my own onboarding into the specific security frameworks we were adopting. I used LLMs to summarize complex compliance documentation and “red-team” my proposed data governance workflows. I use Agentic AI to assist in the multi-step migration of the codebase from a 5 year old beta version to its modern counterpart, saving me weeks, if not months of effort, and allowing me to use that saved time to focus on the roadmap and tech debt prioritization. Utilizing these tools enabled me to rapidly recplaim the domain knowledge we had lost.

Finally, I focused on rebuilding the interpersonal and professional dynamics. Solving the technical problem wasn’t enough if the stakeholders didn’t trust the process. I worked closely with the remaining project managers to establish a regular stakeholder check-in, creating a cadence of transparency that had been missing. We moved from ambiguity to a shared roadmap, re-establishing communication and trust across the institute and with our national and international collaborators.

The result? We successfully secured the first ingestion of controlled-access data in the project’s history. By breaking down the ambiguity into “urgent value” vs. “technical debt” and using transparency to bridge the leadership gap, we didn’t just solve a problem — we stabilized the environment so the team could focus on what they do best: foundational science.