Dhruv Mandalik

Hi, I'm Dhruv, a Computer Science and Electrical & Computer Engineering student at Cornell. I care about efficiency, in models, in code, and in the machines they run on. I enjoy working across robotics, AI/ML, and GPU/low-level programming. In my off hours, you'll find me watching and playing soccer, ping pong, pickleball, exploring new restaurants, and reading.

• InvestcorpSoftware Engineer Intern2026

  Built Python and SQL pipelines that turn unstructured documents into structured, queryable data, combining rule based parsing with an LLM fallback.

• Cornell Tech, AI & Robotics LabML Researcher2026

  Modeled human motion as time series data and benchmarked vision language models, self supervised backbones, and temporal segmentation methods to localize events without labeled data.

• Weill Cornell Meyer Cancer Center, CompBio LabSoftware Researcher2024 – 2025

  Built U-Net and Vision Transformer models in Python to segment organoid microscopy images, reaching 89% IoU and cutting manual analysis time by 75%.

  website ↗

• Virginia Tech, Data Management LabSoftware Engineer Intern2023

  Built vision inference and data logging tools in Python and C++ that helped lab scientists classify nematode species in the field.

Projects

Postwise

GitHub ↗

A privacy focused email agent that runs entirely on device: it categorizes Gmail, answers questions about your inbox, and drafts replies that only send after approval.

(Python, Ollama, Next.js, Gmail API, 2026)

Subtext

GitHub ↗

A study of covert communication between AI agents: a sender hides payloads in ordinary text using acrostic, codebook, and parity schemes while a monitor tries to catch it. The channel lands payloads undetected over 80% of the time.

(Python, Hugging Face Transformers, Qwen2.5, 2026)

cctower

GitHub ↗

A control tower for Claude Code: preflight token checks, a stall watcher, and landing reports that check claimed changes against what actually ran. Fully local.

(Node.js, CLI, lifecycle hooks, 2026)

Free Kick Trajectory Reconstruction

GitHub ↗

Reconstructs 3D soccer ball trajectories from single camera broadcast footage: calibrates the camera from pitch geometry, tracks the ball, then fits drag and Magnus force physics models to resolve depth.

(Python, computer vision, physics modeling, 2025)

Neuro-Runner

GitHub ↗

80 tiny neural nets evolve to clear obstacles, live on this page: no ML frameworks, no training data. Watch them learn below.

(TypeScript, canvas, neuroevolution, 2026)

gen 1

Alive0/0Best0Record0

Brain Tumor Diagnosis

Paper ↗

A benchmark of CNNs, boosting, and multiplicative weight update ensembles for MRI brain tumor diagnosis, reaching 96% accuracy across datasets. Published in the Journal of Student Research.

(TensorFlow, Keras, 2023)

Skills

• Languages: C++, Python, TypeScript, Java, C, SQL
• ML: PyTorch, TensorFlow, CUDA, OpenCV, Computer Vision, LLMs
• Systems: Performance, Concurrency, Embedded, Data Pipelines
• Tools: Git, Linux, AWS, Docker, ROS