Despite high-profile layoffs at major tech companies over the past two years, the roles that matter most to enterprise and mid-market technology organizations remain stubbornly difficult to fill. The candidates exist, but they’re selective — they move fast, and they’re almost never responding to inbound job postings.

They’re being recruited. The organizations who find them are those with either exceptional internal TA teams or strong staffing partnerships. Here are the roles that consistently rank as hardest to hire, and what’s driving the difficulty in each.

These aren’t roles where the talent doesn’t exist — it’s that the talent is scarce, selective, and almost never actively looking. Here’s what’s driving the difficulty in each category and what organizations hiring for these roles should expect.

The candidate pool bifurcates sharply: many people can train a model in a notebook; very few can architect, deploy, monitor, and maintain a production ML system with real load and reliability requirements. The senior production ML engineer — who understands MLOps, model versioning, feature stores, A/B testing infrastructure, and inference optimization — is one of the most competed-for profiles in the market. Expect long searches and strong competing offers for genuine depth at this level.

Cloud migration has been underway for years across most enterprise organizations, but experienced cloud infrastructure and platform engineers consistently outpace supply. Multi-cloud experience — particularly AWS plus Azure or AWS plus GCP — commands significant compensation premiums, and candidates with this background have extraordinary optionality. Platform engineers who can design internal developer platforms that abstract infrastructure complexity for product teams are genuinely scarce and in very high demand.

Security talent has been in shortage for over a decade and the situation has not meaningfully improved. The hardest roles to fill: cloud security engineers (AWS/Azure security architecture, identity, CSPM), SOC analysts at Level 2 and above, penetration testers with web and cloud scope, and security architects who can operate at both technical depth and executive communication levels. Industries with regulatory security requirements — healthcare, financial services, defense — compete especially hard for this talent, creating compensation levels that are difficult for other sectors to match.

This specialty sits at the intersection of hardware and software and is one of the most chronically underserved recruiting challenges. Aerospace, defense, industrial automation, robotics, and semiconductor companies all need embedded engineers — professionals working in C, C++, and sometimes Assembly, who understand real-time operating systems, hardware interfaces, and often safety-critical design constraints. There is no bootcamp pipeline for this skill set. It develops over years of working on physical systems, and geographic constraints compound the difficulty further.

Data engineering has become foundational infrastructure for analytics and AI initiatives, and experienced practitioners are consistently oversubscribed. The specific profile in demand: engineers who understand the modern data stack (Snowflake, dbt, Databricks, Airflow, Spark), can design and maintain reliable pipelines at scale, and have the architectural judgment to make decisions about data quality and platform evolution. Entry-level data engineers are plentiful; senior data engineers who can own an organization’s data infrastructure with genuine independence are a much smaller population.

Junior and mid-level full-stack developers are relatively plentiful. Senior engineers who can design systems, make architectural decisions, mentor junior team members, and ship features independently — with current experience in modern frameworks (React, TypeScript, Node, cloud-native backend patterns) — remain in high demand and command strong compensation. The organizations that struggle most are those offering below-market rates, slow processes, or unclear technical growth paths. Senior full-stack engineers have ample options and make decisions quickly.

Technical PMs — those who genuinely understand software engineering, can manage Agile ceremonies with credibility, and hold delivery teams accountable with both empathy and precision — are a genuinely scarce category. Most PMs can manage process; far fewer can operate as trusted partners to engineering leadership who take them seriously as technical peers. PMs with successful delivery records on complex, multi-team software or infrastructure programs tend to be embedded in long-term engagements and don’t surface on job boards often.

If you’re trying to fill any of these roles through standard internal recruiting or general job boards, plan for significantly longer timelines and more competition than your baseline assumptions would suggest. The candidates you need aren’t waiting to be found — they need to be engaged, developed over time as relationships, and presented with an opportunity that is genuinely compelling compared to their current situation.

PDS maintains active pipelines in all of these categories, with particular depth in IT, AI/ML, data engineering, and embedded systems. Our recruiters work these networks on a continuous basis — not just when a new req arrives. Talk to our team about what’s available in your market.