← Back to All Openings

Principal ML/AI Engineer – Adaptive Differential Privacy & Multi‑Dimensional Reputation Engine

corpora-jobs-1778796293285-db9d41c6 - Frontier Development
ML/AI EngineerPrincipal1 position
⚡

Why This Role is Different

Frontier Development Role

Own the intelligence that turns raw client updates into trustworthy, privacy‑preserving contributions. This role blends advanced privacy theory, Bayesian inference, and federated learning engineering to deliver a robust, adaptive trust system.

The Frontier Element

You will create the first end‑to‑end reputation‑driven DP scheduler that dynamically balances privacy and utility in non‑IID, adversarial federated settings, a breakthrough that has no direct precedent in the literature.

🔬

Project Context

Research Area

Adaptive Differential Privacy Layer (ADPL) and Multi‑Dimensional Reputation Engine (MDRE)

From: Trust‑Aware Federated Aggregation in Multi‑Agent Settings

Why This Role is Critical

ADPL and MDRE are the core of TAFA’s trust‑aware weighting, providing dynamic noise scaling, Bayesian thresholding, and soft exclusion. A dedicated ML/AI lead is required to fuse statistical, cryptographic, and graph‑based techniques into a production‑grade pipeline.

What You Will Build

A reputation engine that extracts multi‑dimensional trust signals, performs Bayesian updates, and outputs continuous reputation scores; an adaptive DP scheduler that modulates noise based on reputation; and integration layers that feed these outputs into the aggregation core.

🛠

Key Responsibilities

  • Design and implement the multi‑dimensional reputation engine that aggregates statistical, temporal, content, and cryptographic signals into a Bayesian posterior.
  • Build the adaptive DP noise scheduler that scales Laplace/Gaussian noise inversely with reputation while maintaining target epsilon guarantees.
  • Develop prototype distillation and contrastive loss pipelines that reduce communication payload without compromising robustness.
  • Integrate the reputation and DP modules with the QRAC and BLTL components, ensuring seamless end‑to‑end operation.
  • Conduct rigorous experiments on real‑world datasets (e.g., UAV telemetry, IoT sensor streams) to validate poisoning resilience and privacy utility.
🎯

Required Skills & Experience

Technical Must-Haves

Differential privacy (DP) theory and implementation

Expert
Designing adaptive noise scaling mechanisms.

Bayesian inference and probabilistic modeling

Expert
Updating multi‑dimensional reputation scores.

Federated learning frameworks (TensorFlow Federated, PySyft, Flower)

Expert
Deploying production‑grade FL pipelines.

Graph neural networks and contrastive learning

Advanced
Implementing FGCLM and prototype distillation.

Statistical anomaly detection and robust aggregation

Advanced
Soft exclusion and Bayesian thresholding.

Experience Requirements

  • 7+ years in machine learning research or production ML engineering.
  • Published work on DP‑FL, reputation systems, or robust aggregation.
  • Hands‑on experience building large‑scale federated learning pipelines.

Education

PhD in Machine Learning, Statistics, or a related field with a focus on privacy or federated learning.

⭐

Preferred Skills

  • Experience with LDP and homomorphic encryption in FL.
  • Knowledge of reinforcement learning policy aggregation.
  • Familiarity with EU AI Act regulatory requirements.
🤝

You Will Thrive Here If...

  • Enjoys rapid experimentation and learning from failure.
  • Comfortable navigating high uncertainty and open research questions.
  • Thrives in a high‑autonomy, high‑impact environment.
📈

Impact & Growth

12-Month Impact

Deliver a reputation‑driven DP engine that cuts poisoning impact by >70% while preserving >90% model utility, validated on a fleet of heterogeneous edge devices within 12 months.

Growth Opportunity

Lead the AI trust research division, extend the reputation engine to zero‑shot policy transfer and cross‑domain policy aggregation, and influence emerging privacy and trust standards.

Ready to Push the Boundaries?

If this sounds like the challenge you have been looking for, we want to hear from you. We value what you can build over where you have been.