Tiana Longjam

Tiana Longjam

Tiana Longjam

Data Science & Mathematics

University of Wisconsin–Madison  ·  Class of December 2026

About

I'm Tiana, a Data Science and Mathematics student at UW–Madison with a Business minor in Finance & Accounting. I work at the intersection of machine learning, mathematics, and real-world impact, from neural PDE solvers to full-stack AI applications.

I care about using data to reduce inequality, whether that's surfacing bias in lending, making speech therapy accessible, or improving mental health access.

University of
Wisconsin–Madison

B.S. Data Science and Mathematics
Minor in Business (Finance & Accounting)
Senior · Graduating May 2027
GPA: 3.63 · Dean's List
Sept 2023 — May 2027
UW

Madison, WI

Experience

Applied AI & Analytics - SharkNinja

Aug 2026 - Present
  • Consolidated calendar, email, Teams, and task tracking into one dashboard, eliminating manual checks across 4 platforms by building and deploying through Cloudflare.
  • Translated stakeholder feedback into implementation requirements across multiple coordinated engineering workstreams.
  • Designed 2 SQL tables and 3 API endpoints to enable multi-tool course tagging, adding validation to block conflicting tool selections.

AI Intern - SAIL

May 2026 - Aug 2026
  • Enabled fully personalized speech therapy for each user by designing a Neo4j Knowledge Graph that tracks session history, pronunciation weak spots, and improvement over time, forming the decision-making backbone that determines which exercises the AI assigns next
  • Closed the feedback loop between speech analysis and exercise delivery by engineering a Python data pipeline that feeds Whisper transcription outputs and CNN pronunciation scores into a graph database, allowing a Contextual Bandit RL agent to automatically reprioritize exercises each session based on where the user is struggling most.
  • Delivered a full-stack AI speech companion by collaborating across a 4-person team, owning the Knowledge Graph, DTW progress scoring, and Exercise Bank while integrating with parallel Whisper ASR, CNN audio analysis, and RAG-grounded LLM layers built by teammates.

Data Science Intern – Infinite Sum Modelling

Jun 2026 – Present
  • Building and developing statistical models using data science and economics tools to support policy and trade analysis
  • Currently being assigned a project to work on

Undergraduate Research Assistant - University of Wisconsin - Madison

Jan 2026 - Present
  • Outperformed classical semi-Lagrangian methods by up to 50x in L∞ error by developing a meshless, GPU-parallelizable neural PDE solver in PyTorch that eliminated exponential growth in degrees of freedom across 8-dimensional problems
  • Reduced computational costs by 10x and L² error by 10x by integrating natural gradient preconditioning and adaptive sampling into a time-sequential neural solver that removes time step stability constraints
  • Validated a novel neural semi-Lagrangian framework through rigorous four-component error analysis and extension to nonlinear Vlasov-Poisson equations, achieving under 1% relative L² error using a second-order predictor-corrector scheme

Community Assistant - University Housing

Nov 2024 - June 2026
  • Utilize software tools including Microsoft Office Suite (Outlook, Word, Excel), SharePoint, and databases to efficiently complete administrative and operational tasks.
  • Leverage spreadsheets and web-based systems to track, organize, and manage information accurately.
  • Demonstrate core values of care, integrity, creativity, respect, optimism, stewardship, and excellence in all interactions.
  • Maintain desk security and confidentiality, and respond promptly to emergencies using StarRez, SharePoint, and EMS

Data Science Intern – Deloitte

Jun–Aug 2024
  • Engineered and consolidated datasets by analyzing confidential client BIG data from over 5 financial years, for a Loyalty Case Study analysis.
  • Contributed to the development of a new retail website by optimizing front-end components with HTML, enhancing the user interface and overall user experience.
  • Conducted extensive data analysis using Python and Pandas library to generate financial conclusions, significantly deepening my analytical skills and understanding of financial data.

Projects

SpeakEasy

Option Pricing: Finite Difference vs. Monte Carlo vs. PINN

github Finance Machine Learning

Python

PyTorch

Black-Scholes

  • Built three independent solvers for the Black-Scholes PDE (finite difference, Monte Carlo simulation, physics-informed neural network) by validating each against the closed-form price, achieving 0.037% and 0.025% error for the finite difference and Monte Carlo methods
  • Diagnosed three separate PINN training failures through loss-component analysis, then corrected them via loss rescaling, input normalization, and strike-focused importance sampling, cutting at-the-money pricing error from 55% to 9.6%
  • Extended the finite difference solver to American options by adding an early-exercise projection step, calculating a $0.51 exercise premium and an $81 exercise boundary at maturity
SpeakEasy

SpeakEasy

github Full Stack NLP / AI Machine Learning

Neo4j

Python

Whisper

CNN

Twilio

RL

  • Designed and implemented a Neo4j Knowledge Graph schema modeling user session history, phoneme-level pronunciation scores, and weak area detection to serve as the personalization backbone for an adaptive speech therapy application
  • Engineered a Python data pipeline that writes Whisper transcription outputs and CNN pronunciation scores to a graph database, enabling a Contextual Bandit RL agent to dynamically adapt exercise selection across 500+ users
  • Collaborated with a 4-person team to deliver a WhatsApp-native AI speech companion integrating Twilio, OpenAI Whisper, Mel Spectrogram + CNN audio analysis, RAG-grounded LLM feedback, and a Neo4j-backed personalization engine
Cloud School Pipeline

Cloud-Based Geospatial School Analytics Pipeline

github Data Engineering Geospatial

GCP

BigQuery

Dataform

GCS

SQL

  • Mapped public school access across all 72 Wisconsin counties by writing BigQuery geospatial SQL queries to analyze 2,000+ schools and identify nearest high schools to every middle school
  • Eliminated manual data integration by orchestrating 3 dependency-managed Dataform pipelines that automatically transform raw Parquet files into geospatial BigQuery tables with county-level joins
  • Enabled scalable statewide education analysis by architecting an end-to-end pipeline across 4 GCP services (GCS, Dataform, BigQuery, GCP VMs), provisioned from scratch on a cloud VM
Competitive Programming Analytics

Competitive Programming Analytics

github Data Engineering Machine Learning

Spark

Hive

PySpark MLlib

Gemini API

  • Reduced query latency by 85% by engineering a distributed analytics pipeline using Apache Spark and Hive to process 8,500+ programming problems and 170,000+ solutions, implementing bucketed tables to eliminate shuffle operations
  • Automated schema-aware prompt construction for deterministic query generation by developing a natural language SQL interface with the Gemini API to convert English queries into executable Spark SQL
  • Built a predictive ML pipeline to estimate missing difficulty scores using PySpark MLlib DecisionTreeRegressor, evaluating 10 tree depths and optimizing model performance through R² score analysis
exaapiproject

PrivacyLens

github Hackathon Full Stack NLP/AI

React

FastAPI

Flask

  • Architected a full-stack privacy audit web application using React, FastAPI, Flask, and Google Gemini 2.5 Flash, enabling users to scan 30+ apps from a single phone screenshot and receive plain-English privacy breakdowns in under 30 seconds
  • Engineered a dual-backend system separating AI processing (Flask) from data and scoring logic (FastAPI), integrating Google Gemini Vision API to automatically detect and audit app icons from uploaded phone screenshots with 90%+ detection accuracy across 37 apps
  • Developed a risk-weighted privacy scoring algorithm using real permission and tracker data sourced from the Exodus Privacy database, normalizing scores across 50+ apps to surface actionable insights for non-technical users at CheeseHacks 2026
exaapiproject

Bias & Source Diversity Checker

github NLP/AI Data Analysis

Python

Exa AI API

HuggingFace

  • Built a tool to analyze search results for source bias and diversity using domain distribution metrics (HHI, Gini, Shannon entropy).
  • Performed sentiment analysis on snippets to detect positivity/negativity trends across domains.
  • Visualized insights with bar charts, pie charts, and heatmaps to highlight dominant sources and overall diversity.
HDFS Pipeline

Fault-Tolerant SQL-HDFS Data Pipeline

Data Engineering

Python

Docker

HDFS

SQL

gRPC

PyArrow

  • Improved query performance and fault tolerance by building a Python gRPC service that cached HDFS Parquet partitions and automatically recovered corrupted files after DataNode failures
  • Reduced data processing overhead for 426K+ mortgage records by implementing a SQL → HDFS pipeline with MySQL, PyArrow, and partition-based Parquet optimization
  • Enabled scalable distributed analytics by deploying a Docker-based ecosystem with MySQL, HDFS, WebHDFS APIs, and gRPC for replicated storage and block-level monitoring
Address Lookup

Distributed Address Lookup System

Data Engineering

gRPC

Docker

LRU Cache

Protobuf

  • Improved system reliability by architecting a fault-tolerant microservices system with 5 Docker containers, implementing gRPC communication, round-robin load balancing across 2 backend replicas, and automatic failover with exponential retry logic
  • Reduced backend load by designing a 3-node HTTP caching layer with LRU eviction policies, achieving cache hits for repeated zipcode queries and enabling degraded operation during complete failures
  • Enabled scalable processing of 50,000+ address records by developing a containerized data service using Protocol Buffers serialization and an automated Docker build pipeline
Weather System

Distributed Weather Data System

Data Engineering

Cassandra

gRPC

Spark

Docker

  • Achieved 99% write availability during node failures by architecting a fault-tolerant gRPC server with a 3-node Cassandra cluster, implementing tunable consistency levels (W=1, R=2) to balance availability and reliability across 200+ Wisconsin weather stations
  • Enabled sub-second querying of time-series weather statistics by designing an automated PySpark pipeline to parse and load 10,000+ temperature records into a partitioned Cassandra schema with compound keys
  • Prevented stale reads by implementing graceful degradation strategies with dual read profiles, quorum-based validation, and prepared statements for consistent fallback behavior
reflect

Reflect

github Hackathon Full Stack

React

Vite

Express

Supabase

Google Maps API

  • Engineered a full-stack mental health web application at MadData 2026 by building 4 core features including a facility finder, mood tracker, AI journal, and streak tracker using React, Vite, and Supabase
  • Integrated Google Maps Distance Matrix and Geocoding APIs into a custom Express backend route to calculate real-time driving distance and straight-line distance across 8,000+ SAMHSA-verified mental health facilities, enabling results to sort automatically by proximity
  • Designed and populated a PostgreSQL database on Supabase by parsing 60+ service codes from the SAMHSA National Directory of Mental Health Treatment Facilities into queryable boolean columns, supporting flexible multi-filter search across 8,319 facilities
wisconsin

Wisconsin Census Population Predictor

github Machine Learning Geospatial

Python

GeoPandas

SQLite

Scikit-learn

  • Engineered and integrated multi-source geospatial, raster, and relational data (GeoJSON, shapefiles, SQLite, and land use raster) into unified DataFrames for predictive modeling.
  • Developed and evaluated multiple regression models (LinearRegression, Pipelines with StandardScaler) to predict Wisconsin county and tract-level population, achieving strong explained variance and model interpretability.
  • Applied feature engineering and visualization using GeoPandas, Matplotlib, and rasterio to map spatial patterns, analyze land-use impact, and assess model performance across regions.
loanspic

Loan Insights & Fair Lending Analysis

github Data Analysis Finance

Python

OOP

Binary Search Trees

Matplotlib

  • Engineered a custom object-oriented framework (Applicant, Loan, Bank classes) to model and analyze 2020 Wisconsin HMDA loan data, enabling structured data handling and extensibility.
  • Implemented a Binary Search Tree (BST) to optimize loan lookups and benchmark performance against traditional iteration, achieving significant efficiency gains.
  • Conducted exploratory and statistical analyses on interest rates, applicant demographics, and lending trends, uncovering insights into potential bias and fairness patterns across banks.
wisconsin

SEC Filing & Server Log Intelligence

github Data Analysis

Python

Pandas

NumPy

Visualization

Data Wrangling

  • Developed a Python data pipeline to parse and analyze SEC EDGAR server logs and company filings, integrating real-world datasets to extract patterns in user access, filing activity, and organizational behavior.
  • Leveraged data analysis libraries (Pandas, NumPy, Matplotlib) to clean, aggregate, and visualize large-scale logs, identifying top IP sources, time-based trends, and state-wise filing distributions.
  • Demonstrated advanced data wrangling and exploratory data analysis (EDA) skills by merging heterogeneous datasets, building insight-driven visualizations, and optimizing parsing efficiency for high-volume text data.
books

Amazon Top 50 Bestsellers

github Data Analysis

Python

Pandas

Seaborn

Matplotlib

  • Performed exploratory data analysis (EDA) on a decade of Amazon bestseller data (2009–2019) to uncover sales trends, genre dominance, and rating distributions using Pandas and NumPy.
  • Visualized key insights such as yearly category performance, author popularity, and pricing correlations through Seaborn and Matplotlib.
  • Applied data cleaning, feature engineering, and statistical analysis techniques to identify patterns influencing book popularity and consumer behavior over time.
loanspic

Stock Price Prediction

github Machine Learning Finance

Python

Machine Learning

Scikit-learn

Matplotlib

Data Visualization

Seaborn

XGBoost

Pandas

NumPy

  • Developed a machine learning pipeline to predict Tesla stock price movement using historical OHLC data; implemented feature engineering, classification models (Logistic Regression, SVM, XGBoost), and performance evaluation with accuracy and ROC metrics.
  • Applied financial data analysis techniques by computing derived features (price spreads, volatility, quarter-end effects) to enhance model interpretability and trading relevance.
  • Conducted exploratory and statistical analyses on interest rates, applicant demographics, and lending trends, uncovering insights into potential bias and fairness patterns across banks.

Extracurriculars

dotData

Jan 2026 – Present
  • Competed in MadData hackathon by collaborating with a team to analyze and model a real-world dataset, delivering data-driven insights within a time-constrained environment
  • Engaged with 10+ industry professionals at speaker and networking events to gain exposure to data science career paths and emerging trends in the field
  • Strengthened technical and analytical skills by attending workshops and study sessions focused on machine learning

Wisconsin RangDe

Sep 2025 – Present
  • Wisconsin’s All Girls Recreational South Asian-Fusion Dance Team!
  • Perform at various university and community events throughout the year.
  • Collaborate with team members to choreograph and execute routines combining Bollywood, Bhangra, Contemporary, and Hip-Hop styles.

Wisconsin Emerging Scholars

Jan 2024 – Dec 2024
  • Earned additional academic credits through co-enrollment in calculus and the WES calculus course, demonstrating commitment and dedication to academic excellence within the WES community.
  • Contributed to a collaborative learning environment as a member of a diverse student group, gaining unique problem-solving techniques and achieving high academic performance, consistently outperforming standard calculus sections.

Skills

Languages
Python
R
SQL
HTML
CSS
Data Science
NumPy
GeoPandas
Matplotlib
rasterio
Pandas
Jupyter
Machine Learning & AI
PyTorch
PySpark MLlib
Scikit-learn
HuggingFace
TensorFlow
NLP
Big Data
Spark
Hive
HDFS
Protobuf
Cassandra
Kafka
Cloud & DBs
BigQuery
Google Cloud Platform
Dataform
Supabase
PostgreSQL
SQLite
Systems
🐳
Docker
LRU Caching
Fault Tolerance
gRPC
microservices
Web & APIs
React
Vite
Express
FastAPI
Flask
Google Maps API
Gemini API
Exa AI API
REST APIs
Tools
Git
Linux/Bash
EMS
Microsoft Excel

Relevant Coursework

Introduction to Artificial Intelligence

Comp Sci 540

Data Science Major

Machine Learning

Knowledge-based search techniques

Automatic deduction

Predicate logic

Probabilistic reasoning

Problem solving

Data mining

Natural language understanding

Computer Vision

Speech recognition

Robotics

Modern Algebra

MATH 541

Mathematics Major

Groups

Normal subgroups

Cayley's theorem

Rings

Ideals

Homomorphisms

Polynomial Rings

Abstract Vector Spaces

Introduction to Finance

Finance 300

Business Minor

Corporate finance and investments

Financial Environment

Securities Markets

Financial Markets

Financial Statements and Analysis

Working Capital Management

Capital Budgeting

Cost of Capital

Dividend Policy

Asset Valuation

Investments

Decision-making under Uncertainty

Mergers

Options