Hi, I'm Leann Alhashishi
Machine Learning Researcher
[resume] · [scholar] · [github] · [linkedin] · leannalhashishi(at)gmail(dot)com
I work on machine learning, with a focus on computer vision and weight-space learning. I just completed an MPhil in Machine Learning and Machine Intelligence at the University of Cambridge. I've previously interned as a software engineer at Cisco and the National Nuclear Security Administration.
Recent projects
Memory-Intensive Tasks with Test-Time Training
ICRA 2027 (submitted). [PDF]
Fit per-episode neural implicit representations over robot video via MAML, then projected the weights as tokens into TinyLlama-1.1B (QA) and Pythia-400M (action control). No raw frames at inference: +19.1pp over baseline on memory-dependent video QA, and R² +0.51-0.54 on held-out joint-action regression.
2026
InLink
Building a weight-space memory that learns an industrial plant's long-term fault patterns. Predicts recurrence 2x better than pooled embeddings and flags novel faults at 0.99 AUROC, where RAG and RAG+SQL score at chance.
2026–present
Upper-Limb Synergy Modeling from Real-World Contactless Motion Tracking
BIOROB 2026 (oral). [IEEE] [PDF]
Reduced synergy dimensionality by ~50% (7-10 to 3-5 dimensions) and captured ~10% more variance than PCA (p < 1e-7 across 40 subjects) using Robust PCA on noisy vision-tracked hand motion, validated on a 6-DoF robotic hand.
2026
Leveling Up the Relationship: Co-designing a Social Biosignal Sharing App to Foster Emotional Awareness and Sense of Belonging
CHI 2027 (submitted).
An HCI co-design study exploring social biosignal-sharing technology recorded over weeks using Empatica’s E4 wristband.
2026
Click here to sign my site.