nahal

nahal🌱

ML X COMP NEURO

Waterloo alum interested in software, building things, machine learning, and computational neuroscience. I’ve worked across software and AI engineering, robotics research, and biosignal systems, and I love building adaptive, real-world systems end to end.

Two sides of the same coin: using AI to study the brain, and using the brain to rethink AI

Education

Sept 2021 — June 2026Graduated

University of Waterloo

Bachelor of Mathematics

Computational Mathematics · Minors in CS & Cognitive Science

Across five years of co-op, research, and interdisciplinary coursework, I became increasingly drawn toward how intelligent systems learn, represent information, and adapt — both in machines and in the brain.

President's Scholarship of DistinctionPresident's Research Award 2026

Experience

Where I've worked, what I've built.

nowAug 2026 – presentFULL-TIME

Amazon

Software Development Engineer

Building software at Amazon.

Feb 2026 – July 2026· CNRG Lab · Waterloo, ONRESEARCH

National Research Council Canada

Student Research Assistant, Robotics & Deep Learning

Taught a skid-steer robot to navigate by learning from expert demonstrations.

Developed an inverse reinforcement learning pipeline to recover navigation cost functions from expert-demonstrated trajectories. Implemented MPC pipelines for real-time, risk-aware trajectory planning, and built a deep learning system that translates hand-drawn paths into structured planner inputs.

inverse RLMPCrisk-aware planningdeep learningroboticsPython

Sept 2025 – Apr 2026· Redwood City, CAINTERNSHIP

Cognichip

ML Engineer, Agentic AI Platform

Built LLM evaluation and agent infrastructure for AI-driven chip design.

Designed and shipped an internal LLM evaluation platform (React, TypeScript, Python, FastAPI) for benchmarking agent behavior, tool reliability, and retrieval quality. Built MCP tools and agent orchestration workflows for multi-step tasks across specialized sub-agents, and improved internal RAG benchmark performance by ~10% by iterating on retrieval, grounding, and embedding pipelines.

LangSmithLangChainMCPFastAPIReactTypeScriptLLM evalRAG

Jan 2026 – Apr 2026LEADERSHIP

UW Data Science Club

VP of Workshops

Established UW Tinkerer Studio, a mentored open-source AI program.

Built out a new collaboration between DSC and HumanFeedback where students form teams to work on real open-source AI projects with mentor guidance.

mentorshipprogram designHumanFeedback collab

June 2025· EVENT

NASA Goddard · Space Apps 2024 Winner's Celebration

Attended NASA Goddard as a Space Apps 2024 finalist.

Jan 2025 – Jan 2026LEADERSHIP

UW Data Science Club

Workshop Lead & Mentor

Ran workshops and mentored students in ML and computational neuroscience.

Led technical workshops including Intro to ML and Intro to Computational Neuroscience. Mentored students across skill levels through their first hands-on projects.

workshop designmentorshipML fundamentalscomp neuro intro

Oct 2024· LEADERSHIP

NASA Space Apps Challenge · Local Lead

Led the local chapter of NASA's annual global hackathon.

Sept 2024· LEADERSHIP

AI Tinkerers Toronto · Co-host

Co-hosted AI Tinkerers meetup at Shopify Toronto.

May 2024 – Dec 2024· Oakville, ONINTERNSHIP

HealthyHer.life

AI/ML Software Engineer Intern

Shipped Hailey, a women's health conversational AI, from concept to deployment.

Led development of Hailey end-to-end. Engineered prompt strategies that improved empathy and accuracy by 15-25%, built a QA logging GUI, and explored LangChain and RAG approaches with Mila Institute advisors.

LangChainRAGconversational AIprompt engineeringMila collab

Sept 2023 – Apr 2024· University of WaterlooDESIGN TEAM

WATOlink

ML Engineer

Biomedical ML project: reconstructing EMG signals into speech.

Worked on the Simphonic's project, building deep learning models (LSTMs and autoencoders) to reconstruct muscle signals into speech output. Contributed to the ML engineering side of cross-modal signal translation for real-time biomedical applications. The project was 1st place winner of CUCAI 2024!

PyTorchLSTMsautoencodersEMGsignal processingbiomedical ML

Sept 2023 – Dec 2023· Markham, ONINTERNSHIP

Assetsoft Consulting

AI Software Engineer Intern

Deployed an invoice classification pipeline used by 500+ enterprise users.

Built a deep learning model for invoice element classification combining NLP and Neural OCR. Deployed model and backend APIs to Azure within .NET infrastructure, and designed a scalable database schema.

deep learningNLPOCRAzureC# .NET

June 2023 – June 2026· LEADERSHIP

Women in AI & Robotics · Student Ambassador

Became a student ambassador for Women in AI & Robotics.

Jan 2023 – Apr 2023· Etobicoke, ONINTERNSHIP

Myant Inc.

Software Engineer Intern

Built real-time biosignal processing for wearable sensors.

Processed ECG, temperature, pressure, and motion signals from wearable sensors. Built a BLE-enabled Python app that replaced a legacy C# interface for real-time acquisition, integrated ML classification models, and improved data efficiency by 30%.

PythonBLEECGbiosignal processingML classification

Research

Research at the intersection of intelligent systems, computational neuroscience, and adaptive behavior — somewhere between a robotics lab and a neuroscience department.

May 2025 – Feb 2026Accepted · CogSci 2026Computational Neuroscience

Basal Ganglia Dynamic Neural Fields

A DNF model of Parkinsonian freezing of gait

Dr. Jeff Orchard, Dr. Madeleine Bartlett · Computational Neuroscience Research Group

Investigated how dopamine depletion alters action-selection dynamics in the basal ganglia through a Dynamic Neural Field (DNF) model of Parkinsonian freezing of gait (FOG). By modeling competitive interactions between direct and indirect pathways, the system reproduced characteristic freezing onset behavior under dopamine loss, providing a population-level dynamical account of impaired motor switching in Parkinson’s disease.

◆Built and tuned a population-coded Dynamic Neural Field model in Python capturing competitive direct/indirect pathway dynamics under healthy and dopamine-depleted conditions.
◆Reproduced empirically observed freezing onset and release behaviors, identifying bistable attractor dynamics as a candidate mechanism underlying FOG.
◆Accepted as a first-author poster at the 2026 Cognitive Science Society Conference (CogSci 2026).

Dynamic Neural FieldsBasal GangliaParkinson'sPythonCogSci 2026

Feb 2026 – July 2026Research ProjectRobotics & Reinforcement Learning

Inverse RL for Risk-Aware Robot Navigation

Learning navigation preferences under uncertainty

Dr. Michael Furlong, Dr. Chris Eliasmith · Computational Neuroscience Research Group

Explores how autonomous systems can infer human-like navigation preferences under uncertainty through inverse reinforcement learning and risk-aware control. Using a skid-steer robotic platform, the project investigates how latent cost functions, risk sensitivity, and adaptive decision-making behaviors can be recovered directly from expert demonstrations.

◆Combines inverse reinforcement learning, Bayesian Optimization, and Model Predictive Control (MPC) for interpretable robot navigation.
◆Focuses on uncertainty-aware planning and recovering latent behavioral preferences from demonstrations.

Inverse RLRisk-Aware ControlMPCRoboticsPython

Winter 2026Course Project · AMATH 382

Dopamine Terminal Dynamics in Parkinson’s Disease · Biophysical modeling of dopamine regulation and neural gain dynamics

Reproduced and extended a biophysical ODE model of dopamine synthesis, release, and reuptake in Parkinson’s disease to study how neural activity modulates dopaminergic dynamics over time. Introduced an activity-dependent Hill-function gain mechanism linking neural firing behavior to dopamine regulation, producing emergent oscillatory behaviors absent from the original model.

Winter 2025Course Project · SYDE 552

Novelty Detection with Legendre Memory Units · Temporal memory architectures for anomaly detection in sequential data

Investigated biologically inspired temporal memory architectures for unsupervised novelty detection in sequential data using Legendre Memory Units (LMUs). By encoding continuous-time history through structured orthogonal memory representations, the project explored how temporal context can support anomaly detection and distribution-shift sensitivity in dynamic environments.

Skills

Languages, frameworks, and tools I work with:

full skills page

Contact

Happy to chat about comp neuro, brain-inspired AI, research, or whatever else:

Or find me on

linkedin.com/in/nahalhz github.com/nahalhz