Case study
71.9% → 100%Melinoe: job application automation
A Playwright-based agent that discovers entry-level postings and auto-fills applications across Greenhouse, Lever, and Ashby, using only the applicant's own resume, templates, and preferences. A form counts as ready only when the resume is attached and every required field is verified filled.
Field success by iteration
Baseline
41/57 fields
Iter 1
50/50 fields
Iter 2–3
53/53 fields
Per-field logging against live forms turned failures into a labeled error set; three fix iterations closed the gap to 100%.
By ATS platform
Two metrics per platform separate parser coverage (filled) from end-to-end completeness (ready): the gap shows where edge cases live.
Application pipeline
- Ready28%
- In progress42%
- Blocked20%
- Unprocessed10%
Every posting carries a tracked state in SQLite, so throughput and blockers are measured, not guessed.
Philemon
Systems that reason, pipelines that hold up.
Philemon is the umbrella for my ML and data science work: agentic systems and applied modeling on real, messy data. Everything here is built end-to-end, from problem framing to deployment.
- live
Melinoe
Job application agent: search, tailor, apply, track.
- LLM orchestration
- Playwright
- Python
View project
- building
Eos
Daily briefing agent: your morning, synthesized.
- Agent pipeline
- RSS/APIs
- Python
- live
Peitho
Conversational lead intake that qualifies while you work.
- Conversational UX
- Typebot
- Next.js
View project
- building
Plutus
Multi-agent financial analysis: theses argued, not summarized.
- Multi-agent
- Cost-optimized inference
- Python
View project
Education
B.S. Data Science, Minor in Computer Science
Rutgers University · June 2026
- Machine Learning
- Data Structures & Algorithms
- Statistical Modeling
- Regression Methods
- Q-Learning
- Applied ML Concepts
Competition
Kaggle
Notebooks and competition work.
Security
Yearly CTFs
Capture-the-flag competition.
- DEFCON Qualifier
- Texas Security Week