TalentPilot

Empowering applicants with a massive, Groq-powered career ecosystem featuring an interactive resume architect, simulated interview coach, and dynamic global tech market analytics.

Comment

Inspiration

Watching friends submit hundreds of job applications with zero feedback was the spark. One friend spent two weeks tailoring a resume, got rejected in 24 hours, and had no idea why. Career coaching costs $200+/hour. Free tools only do one thing — analyze OR build OR match jobs. Nothing connected the full loop. We wanted a single platform where you upload once and get: analysis, fixes, a new resume, interview practice, career advice, job matches, and market trends — all free, all AI-powered.

What it does

TalentPilot is an all-in-one AI career platform. Upload your resume once and get:

ATS Resume Analysis — scored across Content, Formatting, Skills, and Impact with specific fix suggestions Resume Builder — 7-step wizard with 4 professional templates, generates downloadable PNG resumes Resume Refiner — chat with AI to iteratively improve your existing resume Interview Coach — simulated interviews with role-specific questions and answer feedback Expert Reviewer — deep 9-dimension analysis with individual scores and written feedback Career Path Advisor — identifies best-fit careers, alternative roles, and upskilling recommendations Job Recommendations — country-specific listings ranked by resume compatibility Tech Dashboard — country-by-country breakdown of technology demand, salaries, and trends Roadmap Generator — visual career roadmaps for any technology, plus 8 pre-made popular roadmaps Tech News — curated industry news with Edge Computing default feed Every interaction is session-aware — conversations persist across refreshes.

How we built it

Tech stack: Python FastAPI backend, vanilla JavaScript + Tailwind CSS frontend, Groq API for all AI, Pillow for PNG resume generation, Supabase for storage.

The core insight was using Groq's JSON Schema-guided structured output. Instead of parsing free-form LLM text, we defined Pydantic models for every response type (analysis scores, dashboard data, career paths) and passed the JSON schema directly to the API. This forced the LLM to return perfectly typed data that we could validate instantly — no regex parsing, no hallucinated fields.

Challenges we ran into

Prompt engineering. Getting the LLM to consistently output valid nested JSON with exactly 5 technologies, 3 skill areas, and 2+ countries per area took dozens of iterations. Even with JSON schema enforcement, the model would occasionally skip fields or include empty arrays.

Validation cascading. When one AI call returned bad data, it crashed the entire endpoint. We had to add hasattr guards everywhere, relax min_length constraints, and make every field optional with graceful degradation

Accomplishments that we're proud of

10 distinct AI-powered features in a single platform with one resume upload 4 professional resume templates rendered server-side as PNGs with zero paid API costs Structured AI output that's 100% type-safe — every response validated against Pydantic models Session-aware conversations — users can leave and come back without losing context Global tech dashboard with country-level salary/trend data across 8 countries Under 2-second response times for most AI calls thanks to Groq's inference speed Zero frontend build tools — no webpack, no vite, no npm. Just HTML + JS + CDN CSS.

What we learned

SON Schema-guided LLM output is a superpower, but design your schemas defensively — optional fields with defaults everywhere Vanilla JS + Tailwind can absolutely build a polished SPA without a framework A single AI provider handling analysis, generation, chat, and structured extraction reduces integration complexity enormously Pillow image generation is viable for templates you control — just get your font paths right Session-aware conversations are trivial to implement (JSON array per session ID) but feel magical to users

What's next for TalentPilot

User authentication — Google/GitHub login with persistent profiles and resume history Database-backed sessions — replace in-memory storage with PostgreSQL for persistence across devices Real job board integration — connect to LinkedIn/Indeed APIs for live job listings AI-powered cover letter generator — one-click cover letters matched to job descriptions ATS simulator — upload a job description and see exactly how your resume scores against it Mobile app — React Native wrapper for on-the-go resume management

Built With

Share this project:

Updates