Skip to content
View aarshiagupta's full-sized avatar
  • UC San Diego
  • La Jolla, California
  • 07:57 (UTC -12:00)
  • LinkedIn in/aarshiagupta

Highlights

  • Pro

Block or report aarshiagupta

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
aarshiagupta/README.md

Aarshia Gupta

Data Science and Business Analytics @ UC San Diego · Applied AI and Agentic Systems · AI Intern @ Lightship Neuroscience · prev EY

I build LLM pipelines, AI agents, and ML systems that make it to production. Data Science and Business Analytics at UC San Diego, graduating June 2027. Currently AI Intern at Lightship Neuroscience, research assistant at the Rady School of Management under Professor Robert Sanders, and part of Dr. Lara Rangel's Neural Crossroads Lab.

Most of my work sits in AI/ML engineering and applied AI — building agents, pipelines, and models end to end. I'm also drawn to forward-deployed engineering, GTM engineering, and growth, where you sit close to the customer, build the thing, and see whether it actually moved a number.

What I work on

  • Applied AI: agentic systems, retrieval-augmented agents, structured extraction from unstructured text, automated evaluation and scoring pipelines
  • Machine learning: forecasting, classification, geospatial modeling, and model interpretability
  • Causal inference: identification strategy, difference-in-differences, and the endogeneity problem behind pricing models
  • Full stack: FastAPI backends, Next.js and Streamlit frontends, PostgreSQL/pgvector and Supabase, deployed on GCP and AWS

Highlights

  • Lightship Neuroscience: AI Intern building LLM pipelines, retrieval-augmented agents, and automated evaluation systems. Designed a parameterized scoring rubric that replaced hand-tuned weights with measured ones.
  • Rady School of Management: Research under Prof. Robert Sanders on retail pricing and product image data — field data collection, labeling infrastructure, and identification strategy for dynamic pricing.
  • Neural Crossroads Lab: Built an ultrasonic vocalization detection pipeline for behavioral audio analysis, cutting manual annotation time substantially.
  • Moores Cancer Center: Built Python ETL pipelines integrating REDCap, Epic, and FreezerWorks — reshaping flat clinical exports into structured per-aliquot records for biospecimen tracking.
  • EY India: Demand forecasting across 200K+ FMCG products using Random Forest and SARIMA.
  • DS3: Social & Fundraising Director for UCSD's 500+ member Data Science Student Society.
  • UC San Diego: Provost Honors, 2026 Christine Norris Essay Award.

Tech Stack

Languages

Python SQL Java C++ JavaScript

AI / ML / NLP

PyTorch TensorFlow scikit-learn LangChain Anthropic

Data & Cloud

AWS Apache Spark Snowflake PostgreSQL pandas NumPy

Web, Viz & Tools

React D3.js Streamlit Power BI Tableau Git

Core competencies: Agentic AI · RAG · LLMs & Prompt Engineering · Neural Networks · Generative AI · ETL Pipelines · Causal Inference · Hypothesis Testing · Quantitative Analysis · SDLC / UAT

Reach me

LinkedIn UCSD Email


Off the clock: chasing good coffee, logging every restaurant I eat at on Beli, and knowing far too many Bollywood dialogues by heart.

Pinned Loading

  1. kanglee05/Surf-Cast-SD kanglee05/Surf-Cast-SD Public

    Jupyter Notebook

  2. coastal-shield coastal-shield Public

    ML-powered harmful algal bloom early warning system for California coastal health departments - conformal prediction, SHAP explainability, and Gemini-generated audit-ready advisory memos.

    Python 1

  3. luna luna Public

    Interactive sleep-data storytelling site built on the MMASH dataset — take a sleep quiz, find your sleep twin, and explore heart rate, hormone, and lifestyle charts from real participants. Built wi…

    JavaScript

  4. llm-driver-assessment llm-driver-assessment Public

    Retrieval-augmented, tool-calling AI agent (Pydantic AI + Claude) for designing driver-safety assessments, backed by PostgreSQL + pgvector

    HTML

  5. RL-Poker-Bot RL-Poker-Bot Public

    Forked from ChinmayB1/RL-Poker-Bot

    DQN, NFSP, and MCCFR agents for Texas Hold'em — comparing deep RL against game-theoretic approaches to imperfect-information play.

    Jupyter Notebook 1

  6. multi-asset-modeling multi-asset-modeling Public

    Testing whether ES/NQ futures returns lead SOXX due to trading-hours asymmetry, using walk-forward validated logistic/linear regression on 1M+ minute bars.

    Jupyter Notebook