AK@adityan-kesavan·🟢 Open to work
Adityan Kesavan portrait

Third-year Computer Science student building toward a career in tech

I am a 3rd year student studying Computer Science, exploring projects and skills as I prepare to start my career.

Work samples

A collection of class projects and personal work that demonstrate fundamentals in programming, problem solving, and software development as I ready myself for entry-level work.

  • Triagent

    Personal Project · Nov 2026

    Triagent

    I increased incident-severity accuracy from 38% to 93% and correct abstention from 0/7 to 7/7 on 32 labeled incidents by developing a hybrid-retrieval RAG pipeline and LLM agent with code-enforced citations. I enhanced a 1.5B model’s severity accuracy from 28% to 88.5% on 200 incidents through

    PythonFastAPIRAGLLM AgentsQLoRADockerGitHub Actions
  • Vestora
    Vestora

    Built stock research and paper-trading features by integrating 4 REST APIs—Finnhub, Alpaca, Twelve Data, and Gemini—through TypeScript backend services and Supabase Edge Functions. Implemented generative AI assistance with Gemini large language model (LLM) tool calling and confirmation for write act

    React NativeTypeScriptSupabasePostgreSQLREST APIs

About me

I am a 3rd year student studying Computer Science, exploring projects and skills as I prepare to start my career.

I am Adityan Kesavan, a third-year Computer Science student working toward a Bachelor’s degree. I have limited professional experience and am preparing to start my career in software and computing.

Experience

Freelance AI Quality Analyst

Outlier · May 2026 – Nov 2026

Education

Western University

Bachelor of Science in Computer Science · Class of 2024

Skills

Computer Science fundamentalsProgramming (general)Problem solvingCoursework projectsLearning and adaptingAIAI Agent DevelopmentAI Powered AnalyticsAI Workflow DesignAI-Driven InsightsArtificial IntelligenceAutomationBack-end DevelopmentCI/CDClaudeCloud Computing
Back to work

Triagent

I increased incident-severity accuracy from 38% to 93% and correct abstention from 0/7 to 7/7 on 32 labeled incidents by developing a hybrid-retrieval RAG pipeline and LLM agent with code-enforced citations. I enhanced a 1.5B model’s severity accuracy from 28% to 88.5% on 200 incidents through

PythonFastAPIRAGLLM AgentsQLoRADockerGitHub Actions

Overview

I increased incident-severity accuracy from 38% to 93% and correct abstention from 0/7 to 7/7 on 32 labeled incidents by developing a hybrid-retrieval RAG pipeline and LLM agent with code-enforced citations. I enhanced a 1.5B model’s severity accuracy from 28% to 88.5% on 200 incidents through

View all work

Vestora

Built stock research and paper-trading features by integrating 4 REST APIs—Finnhub, Alpaca, Twelve Data, and Gemini—through TypeScript backend services and Supabase Edge Functions. Implemented generative AI assistance with Gemini large language model (LLM) tool calling and confirmation for write act

React NativeTypeScriptSupabasePostgreSQLREST APIs
View all work

Housing Burden Data Pipeline & ML Analysis

Standardized data preparation for 57,936 records and 390 modeling variables by scripting an automated Python ETL pipeline to merge 448-column source data, clean features, and remove data leakage. Achieved a test R2 of 0.705 by training and evaluating Elastic Net and XGBoost machine learning models;

PythonpandasNumPyXGBoost
View all work

DuckSpeak [Demo]

Enabled browser-based ASL recognition from 21 hand landmarks by collaborating on a computer vision application using MediaPipe, normalized feature vectors, KNN, and logistic regression. Applied temporal smoothing and confidence thresholds to live recognition predictions; persisted datasets and model

ReactTypeScriptMediaPipeIndexedDB
View all work