Neha Balamurugan

Systems Engineer / Building Intelligent Systems

Cognitive AI

Mode Select

[GitHub] [LinkedIn] [Email]

About // Profile

Hi, I'm Neha [NAY-ha]. I study how humans and AI systems reason about the world differently, with a focus on visual understanding and social cognition.

Currently pursuing my MS in Computer Science (AI) at Stanford, working in the Causality and Cognition Lab on visual social reasoning benchmarks.

Human AI Alignment Cognitive Science Vision-Language Models Science Integrity

About // Profile

Hi, I'm Neha [NAY-ha]. I build intelligent systems that help people equitably reason, learn, and create.

I love building AI tools that solve real problems.

AI Systems Fintech Femtech VR/AR

Experience // Timeline

PRESENT

</building>

Stealth

2025

Emerson Collective // Venture

Explored the software-for-hardware market

2023

Checkbook // Software Engineer

Built fintech infrastructure for digital payments

2022 - 2024

Dreamscape Learn // VR Engineer

Led team building multiplayer VR classrooms deployed to 1,500+ students

Academic Profile // Timeline

2024 - 2026

Stanford University // MS Computer Science (AI)

Causality and Cognition Lab / Accel Leadership Scholar

- Developing benchmarks to study visual social reasoning in humans and AI

- 1 of 16 students selected for Stanford's flagship entrepreneurship program (ALP)

2021 - 2024

Arizona State University // BS Computer Science

RANK 1 // CS UNDERGRAD

- Led 10-member team under $500M Dreamscape Learn VR initiative

- DAAD RISE Fellow at Humboldt University Berlin

- Multiple hackathon wins across education, fintech, and health tech

Primary Focus Vision-Language Models

Affiliation Stanford HAI

Recognition HLF 2025 Scholar

Primary Focus AI Systems

Affiliation Stanford University

Scan_ID: NB-001

Research Archive // Publications

ID: RES-2025-001

Spot The Ball: Visual Social Reasoning Benchmark

Developed a benchmark measuring how humans and AI infer hidden information in sports scenes. Evaluated frontier VLMs to study human-AI reasoning gaps.

VLM COGNITIVE

[ArXiv]

ID: RES-2024-002

Dreamscape Learn VR Education

Led 10-member team under $500M initiative to create multiplayer VR classrooms. Built interactive visualizations deployed to 1,500+ students.

VR UNITY

[ILRN 2024]

GCSP AWARD 5X

ID: RES-2023-003

AR Driver Safety System

DAAD RISE Fellowship: Designed AR heads-up displays to reduce driver distraction using CARLA simulations, eye-tracking, and HoloLens experiments.

AR HOLOLENS

BARRETT HALL OF FAME

ID: RES-2024-004

buff: AI for Scientific Research

Undergraduate thesis: AI framework combining LLMs, citation graph analysis, and modular retrieval pipelines to accelerate scientific discovery.

LLM RAG

ID: RES-2022-005

Fluid Dynamics Visualization in VR

NSF REU: Built interactive VR systems for exploring large-scale fluid simulations, improving real-time rendering speed by 45%.

VR GPU

Engineering Archive // Systems Built

HACK: CALHACKS-25

Recall: Wearable Conversation Recorder

Raspberry Pi wearable with voice detection, FaceNet embeddings, Gemini transcription, and Supabase storage for natural-language search.

HARDWARE AI

[Devpost]

HACK: TREEHACKS-23

YouCare: Women's Wellness Forecasting

iOS app for personalized health analysis using Apple HealthKit and You.com AI search platform for women's wellness tracking.

SWIFT iOS

[Devpost]

BEST OPEN PLATFORM

HACK: HACKMIT-23

SusProduce: Food Usability AI

ML image detection using Keras to assess food usability past expiration. Angular frontend for intelligent food waste reduction.

ML KERAS

BEST SOLUTION CPG

HACK: H4H-23

CareUp: Disability Caretaker Assistant

AI-powered app that finds, compiles, and ranks resources for elderly/disordered family members and assists with form completion.

AI HEALTH

1ST PLACE

HACK: TREEHACKS-24

Althea: AI Research Agent

Research assistant for literature reviews fine-tuned on ~500 biochemistry papers with citation network integration and semantic chunking.

LLM RAG

[Devpost]

Conference_Log // Engagements

Ref_ID	Venue	Year	Engagement_Type	Topic_Node
CONF-COGSCI-26	CogSci // Cognitive Science Society	2026	Presenter	Visual Social Reasoning in VLMs
CONF-HLF-25	HLF // Heidelberg Laureate Forum	2025	Scholar	Cross-disciplinary AI Ethics
CONF-ILRN-24	iLRN // Immersive Learning Research Network	2024	Publication	VR Education Systems
CONF-DAAD-23	DAAD RISE // German Academic Exchange	2023	Fellow	AR Driver Safety HCI
CONF-FORGE-24	ASU Fulton Forge // Research Symposium	2022-24	Poster	Dreamscape VR Research

Course Projects // Academic Work

CS238 // FALL 2024

Lunar Surface Navigation under Low Visibility

QMDP-based decision-making for lunar rover navigation, achieving 100% success in dense environments.

RL POMDP

NLP // WINTER 2025

Temporal Sense of Large Language Models

Studied how LLMs represent time using modified TOT-Semantic benchmark with linguistic noise.

LLM TEMPORAL

GRAPHICS // WINTER 2025

Volumetric Rendering in Ray Tracer

Physically based volumetric rendering simulating absorption, scattering in fog/smoke media.

GRAPHICS RAYTRACING

DEEP RL // SPRING 2025

Test-Time Prompt Compression for SLMs

Hierarchical framework using Gemma 12B for planning and Qwen2.5-0.5B for generation.

RAG RL

VCS // SPRING 2025

DSPy: ContextSeeker Module

DSPy module enabling LLMs to ask clarifying follow-up questions for underspecified tasks.

DSPY HITL

GNN // FALL 2025

Aircraft Fuel Efficiency with GNNs

Graph Neural Network predicting fuel consumption using spatiotemporal flight trajectory data.

GNN AVIATION

AI4OPT // FALL 2025

Primal Heuristics for NN Verification

Differentiable integer linear optimization for neural network verification without supervision.

MILP NNV

Content // Media & Writing

podcasts

Second Order

Podcast

Mental models for entrepreneurship & leadership. Guests include Elad Gil.

[Listen Now] edit_note

Thoughts & Essays

Substack

Essays on technology, startups, and the intersection of AI and society.

[Read Now]