LLM-Powered Job Application Agent
Applying to jobs can be repetitive and time-consuming. I helped to build an agent that automates large parts of the application process using LLMs, information retrieval techniques, and browser automation. The system retrieves and ranks job postings, understands user queries, and can even submit applications automatically.
Key Skills Demonstrated
- Collected and processed job listings from Lever to create a structured job board dataset for model input.
- Converted job postings into text documents to support retrieval-augmented generation (RAG) pipelines.
- Built a hybrid ranking system using TF-IDF for keyword matching followed by neural re-ranking with BERT for semantic scoring.
- Integrated Meta Llama 3.1-8B Instruct model to interpret user prompts and surface relevant job recommendations.
- Implemented Selenium-based automation to simulate application submissions directly through job board interfaces.
- Handled complex search queries like "Find me ML jobs" or "Search for Apex roles" with multi-step reasoning and retrieval logic.
Project Insights & Learnings
This project allowed me to combine retrieval methods, language models, and automation into one end-to-end system. I had to solve challenges around ranking relevance, handling ambiguity in user prompts, and making sure the agent could operate reliably across different job formats and search types.
Building the application automation layer with Selenium also helped me think more carefully about system robustness, failure handling, and making sure that each automation step was predictable and reversible.