Machine Learning Engineering Intern
Summer 2026ML EngineeringInternship
Lab126: Next-gen consumer electronics R&D
PythonML SystemsAWS
Ishan Dave
Experience
Industry and Research Work
Industry
3Software Engineering Intern
Dec 2025 — PresentFull-StackInternship
Building multilingual product surfaces and self-registration flows for the NanoSence AI platform, scaling to 2,000+ global users across Japanese and English.
Flaski18nSQLJavaScript
Software Engineering Intern
May 2025 — Aug 2025Full-StackCloud
Built cloud-based React + FastAPI demos on AWS and improved backend latency by ~20% by integrating optimized C++ modules with microservices.
ReactFastAPIAWSC++
Research
4Undergraduate Researcher
Oct 2025 — PresentResearchMLBiotech
Designing novel ASO creation mechanisms using ML and evolutionary search.
PythonMLOptimizationResearch
AI Researcher
Jan 2026 — PresentResearchNLPTrustworthy AI
Built an evaluation framework to quantify AI hallucination risk across four academic domains.
PythonBERTNLIEvaluation
ML Research Intern
May 2024 — Oct 2024ResearchVLMsInterpretability
Built interpretability pipelines for vision-language models, published at ICLR 2026.
PythonPyTorchVLMsEvaluation
CV / ML Researcher
Jun 2023 — May 2024ResearchComputer VisionHealthcare
Trained MobileNetV2 on hand drawings to detect Parkinson's at 91% accuracy and 0.92 precision. Placed 3rd at ISEF qualifier, outperforming ResNet50 baselines.
TensorFlowMobileNetV2CVHealthcare
Publications
Peer-reviewed research
ICLR 2026
Visuals Lie, Consistency Speaks: Disentangling Spatial Attention from Reliability in Vision-Language Models
Logan Mann, Yi Xia, Ajit Saravanan, Ishan Dave, Saadullah Ismail, Shikhar Shiromani, Emily Huang, Ruizhe Li, Kevin Zhu
We demonstrate that spatial attention in VLMs is uncorrelated with accuracy; instead, true reliability is encoded in late-layer hidden states and generation consistency.
Read on OpenReview↗
3rd @ SCVSEFA
Automated Detection of Parkinson's Disease Patterns in Spiral Drawings Using Convolutional Neural Networks
Ishan Dave
MobileNetV2 achieves 91% accuracy on digitized spiral drawings for Parkinson's detection, outperforming ResNet50 — offering a non-invasive, cost-effective diagnostic alternative.
Read Paper↗
Projects
Selected personal and academic builds
IceBreaker
Academic Outreach Platform
Automated professor outreach for 40+ students with FastAPI, SQL, and Gmail API integration on Linux.
FastAPISQLGmail APILinux
$
Aionix
Senior Digital Literacy Nonprofit
Built a web platform posting senior digital-literacy webinars — 1000+ users, $16k raised, REST APIs and CI/CD.
JavaScriptAWSRESTCI/CD
$
Current Focus
ML & AI Research
Working on VLMs for robotics with a focus on behavior under distribution shift. Tied to an ICLR-track submission. Also exploring interpretability and trustworthy AI across modalities.
PyTorchVLMsICLRResearch
ML Engineering & Software
Taking models from research to production. Interning at Amazon Lab126 this summer, building ML systems for next-generation consumer hardware.
PythonAWSC++Systems
Physical AI
Interested in the sensing, dynamics, and control problems that come with real-world deployment — sim-to-real, contact-rich environments, and closing the gap between what works in simulation and what works on a table.
RoboticsControlEmbodied AI
Technical Stack
Languages
- Python
- C
- C++
- Java
- JavaScript
- SQL
ML / AI
- PyTorch
- TensorFlow
- HuggingFace
Web / Tools
- React
- FastAPI
- Git
- Linux
Cloud / Infra
- AWS (Lambda, Bedrock)
- Docker
Education
UC Berkeley
Computer Science + Applied Math
What I've been taking
Beyond Classes
Nova Consulting
Consultant
Berkeley technical and strategy consulting—client-facing projects across software and analytics.
Rec Basketball
Sharpshooter
Sharpshooter at the rec during my free evenings.