If you’re a fresher, you know the loop 🔁: another course, another certificate, another 50 applications, silence.
I’m a student, and I lived that loop. So I started building Antenomic, a tool that tells you what real jobs ask for and helps you prove you can do it. It’s in beta now, so here’s how it works and what I learned building it.
🧭 Step 1: Read the market, not a syllabus
The app reads 20,000+ live job posts from company job boards. For your target role it finds the skills you’re missing, but it counts each company once. Early on, one company posting dozens of near-identical jobs was skewing everything.
🗺️ Step 2: Let the AI choose, but never trust it blindly
An LLM picks 4–10 skills and orders them for a fresher. It only sees your data and a fixed candidate list, and anything it invents gets dropped. Lesson: in early tests the model left out 5 of 13 must-have skills, so the server now always puts must-haves back on the path.
🎥 Step 3: Ranking free videos (YouTube is weird)
Searching “Gin tutorial” (Gin is a Go framework) returned… gin and tonic recipes 🍸 and Gin Rummy.
Searching “DBT” returned India’s bank-transfer scheme 🏦.
So there are two layers. An AI check decides if a video teaches the skill, and a plain formula ranks the survivors. Chapters weigh most, because they make the next step possible. When results are thin, the AI writes one clarified search (“Gin Go web framework tutorial”), and the results go through the same check.
🛠️ Step 4: Guided projects pinned to the exact minute
Each skill gets a small project in 4–6 steps, and every step links to the timestamp of a chapter that teaches it. The projects are AI-written, so the server is strict: it refuses a draft that cites a video it wasn’t given, a chapter that doesn’t exist, or a link.
🤖 Step 5: AI code review, with guardrails
You upload your project and get feedback per step: ✅ meets, 🔧 needs changes, plus a hint (never the full solution). Before anything reaches the AI, files are scanned for secrets 🔐, because freshers will upload their .env.
🗑️ What I deleted
I had built an AI “readiness interview”. I removed it. One chat can’t judge whether someone is job-ready, and calling it that was dishonest.
🧱 Stack
FastAPI · SQLModel · React + Vite · SQLite (WAL) · one VM behind Cloudflare · open-weights LLM (~₹0.02 per call).
🙏 I need your feedback
It’s beta and rough in places. Try it at antenomic.com, and tell me in the comments:
What confused you?
Would a reviewed project help you get noticed?
Next up: interview practice and 2-week build sprints 🏁
Top comments (1)
The LLM dropping 5 of 13 must-haves is a good reason to show candidates the extracted list so they can correct it. On your question, a reviewed project helps most when the review says what the fresher decided and why, because that's what a hiring manager can judge in a two-minute read.