Hello! 👋

I'm Ekansh Mittal

Resume

About Me

I am a student at Stanford University, studying CS, AI, and Math. I am interested in the intersection of AI, Robotics, and Autonomous Systems.

I am currently interning at Serve Robotics where I am leading the development of video models for impact detection. I have previously interned at Biostate.ai (spun out into K-Dense.ai), and done research at Stanford University, Beth Israel Deaconess Medical Center, and the Oregon Health & Science University.

Python C++ JavaScript PyTorch MySQL

Experience

Autonomy Intern

Serve Robotics

03/2026 – Present
  • Led training and productionization of an on-device V-JEPA-2 style model across 2,000+ robots to detect delivery impacts, tripling remote supervisor capacity.
  • Designed an automated VLM-based video labeling pipeline, scaling training data from hundreds of samples to production-scale datasets.

AI Intern

Biostate.ai

04/2025 – 10/2025
  • Built MetaMuse, an agentic metadata curation pipeline — see Projects.
  • Designed REVIVE-Flow, a flow-matching model that maps and reverses DNA methylation aging trajectories to propose minimal, biologically plausible rejuvenation edits.

Research Science Institute Scholar

MIT & Beth Israel Deaconess Medical Center

06/2023 – 08/2023
  • Selected among the top 100 STEM students worldwide for a 6-week research internship hosted by MIT.
  • Conducted metabolic rate prediction research using neural networks

Research Intern

Oregon Health & Science University

09/2021 – 08/2025
  • Conducted breast cancer drug resistance research using ML and 3D spheroid cultures — see Projects.
  • Named among Top 40 finalists in the Regeneron Science Talent Search (STS), received $25,000 scholarship.
  • 2nd Place grand award in Cellular and Molecular Biology at Regeneron ISEF; 3× ISEF qualifier (2022–2024).

Projects

MetaMuse: Agentic Curation of Unstructured Biomedical Text

01
  • LLM pipeline extracting structured metadata from biomedical text
  • Saved $1.2M in curation costs, accelerated throughput 450×
  • Unlocked millions of samples for foundation model training
PythonLLMsAgents-SDK

Probing Physical Realism in Diffusion Image Generators

02
  • Trained linear probes on FLUX.1-schnell internal activations
  • Classified physically plausible vs. implausible images across 3 violation types
  • Achieved AUROC 1.000 in-distribution; visual probes beat text baseline on reflection
PyTorchDiffusion ModelsSkLearn

Subset-Consistent Latent Reconstruction for Robust RGB-D Segmentation

03
  • SegFormer-B2 fusion model with teacher-student latent restoration
  • Handles degraded or missing RGB-D sensor inputs
  • Improved average robust mIoU from 0.3996 to 0.4101 on NYUv2
PyTorchSegFormerComputer Vision

Spartan: Edge AI Situational Helmet for First Responders

04
  • Edge AI helmet with on-device VLM inference on Jetson Orin Nano
  • Live dual-eye HUD with night vision and GPS/IMU localization
  • Real-time physiological monitoring (heart rate and temperature)
JetsonVLMsComputer Vision

Sequential RL on Economic Discourse for Real-Time Financial Trading

05
  • Multimodal PPO agent trading on live earnings call transcripts
  • Synced with high-frequency stock data for real-time decisions
  • Longformer yielded higher upside (>4.6% max return) vs. static baselines
PyTorchOpenAI GymAWS

SkyBites: Real-Time Drone Coordination for Scalable Campus Food Delivery

06
  • PPO-based RL coordinating 25 autonomous drones for campus food delivery
  • Achieved 50% reduction in delivery delay over greedy baselines
  • Used Action Masking and Curriculum Learning for complex constraints
PythonLLMsAgents-SDK

Harnessing Machine Learning and 3D Spheroid Cultures to Identify Biomarkers for Combating Drug Resistance

07
  • Neural network with Bayesian Optimization at 92% accuracy
  • Identified IGF1R as key resistance marker from >16,000 gene features
  • Validated triple-therapy with 3D spheroid co-culture; growth reduced 3×
TensorflowSkLearnPandas

Education

Stanford University

  • B.S. Computer Science, Artificial Intelligence Track
  • Minor in Mathematics
2024 - 2028
  • Relevant Courses
    • CS231N — Deep Learning for Computer Vision
    • CS224R — Deep Reinforcement Learning
    • CS137A — Robot Autonomy
    • CS238 — Decision Making Under Uncertainty
    • CS229 — Machine Learning
    • CS205L — Continuous Math for ML
    • CS107 — Computer Systems
    • CS153 — Infrastructure at Scale
    • MATH51 — Linear Algebra & Multivariable Calculus
    • MATH53 — Differential Equations
    • ECON50Q — Quantitative Economic Analysis