AI Job Recommender & Resume Analyzer
Semantic job matching with embeddings, skill-gap analysis and an AI-generated learning roadmap.
I'm Pranjal, a CS undergrad in Lucknow building AI systems that do real work: agents that research papers, a matcher that finds the right job, and a triage desk that decides who sees a doctor first.
Scramble: a loose research question and thousands of papers scattered across sources.
Solve: a CrewAI crew where each agent has one job. One plans the search, one pulls papers from Semantic Scholar, one embeds them into ChromaDB, and one writes a cited summary with LLaMA 3.3 70B on Groq.
PROJECTS list at the top of the file (or later, in each project's .md file) — and replace all of them before the site goes public.Each card is generated from one project file. Add a file and a card appears; set draft: true and it disappears.
Semantic job matching with embeddings, skill-gap analysis and an AI-generated learning roadmap.
Predicts likely conditions from symptoms, ranks severity and issues doctor tokens so urgent cases go first.
What you built in XX hours with a team of X, and your part in it.
Copy _template.md, fill in the fields, and your new project appears here with its own case-study page.
Get it running end to end, then make each step good.
Small, focused steps are easier to test and swap out.
If I can't put a number on it, I can't tell if it improved.
If people can't click it, it didn't happen.
No internship yet, so here are the places I've been trusted with real responsibility.
What the challenge was, what we built, and why it won. details
Ran pre-placement training and related activities for XXX students.
Helped XX juniors set academic goals and build a roadmap to reach them.
Helped run XX meetups and events for the open-source community.
The link between XX students and faculty.
Before that: City Montessori School, Lucknow, Class XII, 93.25%.
I got into AI the same way I got into cubing: something looked impossible, so I broke it into steps until it wasn't. rewrite in your words
Right now I'm most interested in agents that plan and use tools, and in making them reliable enough to trust with real work.
PB 29s. Still chasing sub-25.
Rhythm first. Then everything else.
Four strings, questionable singing.
The fastest route is email. I reply within a day.
I'm Pranjal Verma, a CS undergrad building multi-agent and LLM systems with CrewAI, LangGraph and FastAPI. Open to GenAI and SDE roles, graduating 2027.
view runs ↓resumegithubEvery project is logged as a run: its pipeline, its status, and what came out the other end.
I build AI systems that research, match and triage. I beatbox between builds.
Click a track to open it.
The fastest route is email. I reply within a day.
Email meLinkedInResume