Hi, my name is

Beauty Tasara.

I'm a Software Engineer

I build production software and applied AI systems, from offline field apps and document pipelines to LLM agents and evaluations. I'm finishing an M.S. in Computer Science with a machine learning specialization at Stevens and I'm available to start in February 2027.

Say Hello

01.About Me

I'm a software engineer completing an M.S. in Computer Science with a machine learning specialization at Stevens Institute of Technology, graduating in December 2026. Before graduate school I worked full time as a software developer at Hail & Cotton, after engineering internships at Altonium and Earth Digital.

My work spans full stack development and applied AI. I've shipped an offline first Flutter app used by about 500 field staff, an OCR pipeline that turned tobacco sales sheets into structured records, and a leaf grading model deployed in a mobile app. Lately I've been building on LLM APIs: a purchasing agent kept in check by deterministic rules, a retrieval augmented email assistant, and a team study of positional bias in LLM summaries.

Outside of work and research, you can usually find me at the gym, planning meals, diving into technical papers, or working on ideas that connect AI with everyday life.

Here are a few technologies I've been working with recently:

  • ▹Python
  • ▹TypeScript
  • ▹React
  • ▹Next.js
  • ▹TensorFlow
  • ▹LangChain
  • ▹MATLAB
  • ▹AWS
  • ▹Kubernetes
  • ▹Claude Code
Profile Image

02.Where I've Worked

Graduate Admissions Assistant @ Stevens Institute of Technology

Jan 2025 to Present

  • ▹Handle front desk visitors, admissions email, outreach calls, document filing, and application review support for the Graduate Admissions office.

03.Some Things I've Built

Spenddat, Autonomous AI Purchasing Agent preview
Nov 2, 2026

Featured Project · Personal project

Spenddat, Autonomous AI Purchasing Agent

An LLM agent that turns plain English shopping requests into purchases. Requests within policy run automatically; risky or over budget ones go to a human for approval. Spending limits live outside the model in a deterministic rules engine with hard ceilings no approval can override, and payments use capped, expiring Stripe agent tokens, so the agent never sees card numbers. Includes a tamper evident audit log, a dashboard and chat interface, and 208 automated tests with CI.
  • Python
  • Claude API
  • Stripe
  • SQLite
  • GitHub Actions
Positional Bias Analysis with CoT & Prompt Repetition preview
Apr 14, 2026

Featured Project · Graduate NLP course project, team lead

Positional Bias Analysis with CoT & Prompt Repetition

Led a team testing whether the position of a negative review changes how LLMs summarize mixed opinions. The team built a Yelp based dataset of 100 businesses, each with one negative review placed at the start, middle, or end, then compared chain of thought prompting and prompt repetition across GPT-4o, Claude, and Gemini, scoring summaries with G-Eval, VADER sentiment, and cosine similarity.
  • Python
  • OpenAI API
  • OpenRouter
  • G-Eval
  • VADER
TinySearch, News Recommender preview
Jan 20, 2026

Featured Project · Personal project

TinySearch, News Recommender

A news recommender with a Flask REST API, JWT authentication, SQLite, and a React and TypeScript frontend. The ranker turns likes, dwell time, and clicks into long term and recent taste profiles with time decay, scores articles with local TF-IDF and Truncated SVD vectors instead of a paid embedding API, and serves the feed through stable cursor pagination.
  • Python
  • TypeScript
  • Flask
  • SQLite
  • scikit-learn
  • React
  • Tailwind CSS
G
Jul 28, 2025

Featured Project · RAG prototype

Graduate Admissions Email Assistant

A retrieval augmented service that drafts replies to graduate admissions emails from university web pages and PDFs. Sources are chunked and embedded with a sentence transformer model into a FAISS index, with a fallback to site restricted web search when retrieval finds too little context. Replies come from Claude through OpenRouter and are served through a Flask API.
  • Python
  • LangChain
  • FAISS
  • Hugging Face
  • Flask
ClinicOps, Clinic Management Microservices preview
Nov 22, 2025

Featured Project · Graduate course project

ClinicOps, Clinic Management Microservices

A Java microservices system for patients, providers, and treatments with a REST API, a server rendered web UI, and a backend service that is the only component allowed to reach PostgreSQL. Each service runs in Docker on AWS EKS with Kubernetes Deployments and Services, Amazon RDS, and private ECR repositories, and the backend exposes liveness and readiness health endpoints.
  • Java
  • Quarkus
  • PostgreSQL
  • Docker
  • Kubernetes
  • AWS EKS
Autonomous Path Planning & Obstacle Avoidance preview
Apr 10, 2026

Featured Project · Graduate course project, team lead

Autonomous Path Planning & Obstacle Avoidance

Led a team that built differential drive robot behavior in a simulated 5 m by 5 m arena using an 11 state hierarchical state machine, RGB color detection, and odometry based navigation. The robot completed all six color permutation trials with 12 of 12 deposit points and zero collisions.
  • MATLAB
  • Simulink
  • Stateflow
CNN From Scratch preview
Feb 21, 2022

Featured Project · Personal project

CNN From Scratch

Implemented convolutional, dense, and activation layers with forward and backward passes and cross entropy loss without a deep learning framework, then trained the network on a subset of MNIST.
  • Python
  • NumPy
  • SciPy
View the Archive

04.Get In Touch

Let's talk about everything and anything.

Say HelloGitHubLinkedIn