Varun N.

Key Skills

Software

Other

Top Subject Matter

Meteorological/UAV sensing and drone design optimization

Autonomous driving perception and simulation (cone detection)

Business analytics and decision support for small businesses

Top Data Types

Text

Image

Top Task Types

Function Calling

Segmentation

Freelancer Overview

Research Fellow, ATMOS Lab (AI-generated actionable insights using GPT-4o). Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Science, University of California, Santa Cruz (2029). AI-training focus includes data types such as Text, Computer Code, and Programming and labeling workflows including Function Calling, Segmentation, and Translation.

Labeling Experience

Research Fellow, ATMOS Lab (AI-generated actionable insights using GPT-4o)

OtherTextFunction Calling

Developed an AI-assisted insight generation pipeline using GPT-4o by transforming preprocessed meteorological data into actionable outputs for research decision-making. Used Python and pandas to prepare iMet-X4 (meteorological) inputs and pass them to the GPT-4o API for downstream insight generation. Collaborated with a graduate student to align generated insights with Professor goals for maximizing data collection and improving drone design. • Performed data preprocessing for iMet-X4 meteorological datasets. • Applied GPT-4o API calls to produce actionable narrative insights. • Iterated on signal processing approaches including Savitzky–Golay filtering for UAV vibration denoising. • Partnered with faculty/students to validate insight usefulness for improving drone design and collection strategies.

2026 - Present

Simulations Member - University of California, Santa Cruz

ImageSegmentation

As a Simulations Member at Formula Slug, you run and visualize engineering simulations to support mechanical design improvements. You work with aerodynamic and braking-related performance modeling using industry tools and plotting libraries to communicate results. You also collaborate with a team to optimize computer vision inference for a driverless racing segment using embedded deployment frameworks. • Simulated brake performance in Ansys Fluent and visualized results in Matplotlib • Produced analysis to inform design improvements for the mechanical subteam • Collaborated with a team of engineers to fine-tune a Yolact-Edge segmentation model • Deployed and optimized the model using NVIDIA Jetson AGX with TensorRT for cone detection

2025 - Present

EV Battery and Price Modeling — AI regression and deployment pipeline

Other

Built an end-to-end AI project to analyze and predict EV battery and vehicle prices using machine learning regression modeling. Collected and explored large-scale tabular datasets using pandas, NumPy, and web scraping, then trained a PyTorch regression model to estimate prices from battery features. Designed an API backend for deployment using FastAPI to operationalize the prediction model. • Collected and explored datasets from Kaggle, public sources, and web scraping. • Trained a PyTorch regression model to predict EV prices from battery features. • Evaluated model performance using mean absolute error. • Implemented a FastAPI backend design for public deployment.

2025 - Present

Formula Slug Simulations Member (segmentation model fine-tuning for cone detection)

OtherSegmentation

Fine-tuned a Yolact-Edge segmentation model to improve cone detection for a driverless racing use case. Used TensorRT on the NVIDIA Jetson AGX to optimize inference performance for the Michigan 2026 racing segment. Contributed to training workflow and model refinement as part of a larger team effort to enhance perception accuracy. • Fine-tuned segmentation architecture (Yolact-Edge) for cone detection. • Employed TensorRT optimization for deployment on NVIDIA Jetson AGX. • Produced and reviewed visualization outputs in Matplotlib to support design decisions. • Collaborated with a team of 6 to iteratively improve model performance for the racing scenario.

2025 - Present

Co-Creator, Harbor (CruzHacks2026) — LLM translation of business data to insights

OtherText

Integrated an LLM API into a business analytics workflow to translate structured business data into plain-language insights and recommendations. Called DeepSEEK R1’s LLM API to convert outputs such as tourist demand signals and financial impact estimates into human-readable guidance for small businesses. Focused on producing actionable, conversational explanations suitable for end-user decision-making. • Used DeepSEEK R1 LLM API for plain-language insight generation. • Converted predictive outputs (e.g., demand/traffic, rent impact) into actionable recommendations. • Collaborated with a small team to integrate the AI layer into a full-stack system. • Supported model-building efforts alongside the LLM translation layer.

2026 - 2026

Education

Work History