Bhavesh Solminde - AI Engineer | ContraWork by Bhavesh Solminde
Bhavesh Solminde

Bhavesh Solminde

Full-stack & AI engineer building production-grade apps.

New to Contra

Bhavesh is building their profile!

Cover image for Resumer is a full-stack platform
Resumer is a full-stack platform I built to fix the two most annoying parts of applying to jobs: not knowing if your resume will actually pass an ATS, and rewriting it for every single posting. I built it as a personal project and it's now indexed on Google, live, and has run 300+ resume scans for 100+ users. The core is an AI analysis engine, upload a PDF, get an ATS compatibility score, plus structured feedback on missing keywords, formatting issues, and specific tips to fix them. From there, the optimization engine can rewrite sections generally or tailor them to a specific job description, with a before/after diff view so you can see exactly what changed and why, instead of trusting a black box. I also built a full drag-and-drop resume builder from scratch, multiple templates, undo/redo history, real-time PDF export, so users aren't just getting feedback, they can act on it in the same place. Auth runs on JWT with Google/GitHub OAuth, resumes are stored on Cloudinary with auto-generated thumbnails, and the whole thing is a TypeScript monorepo (React/Vite frontend, Express backend) so the types stay in sync across both sides. The result is a production tool people actually return to, not a one-off demo, scan history, multiple resume versions, and a builder that holds up under real usage.
1
69
Cover image for The client wanted a complete
The client wanted a complete food delivery platform, a customer-facing Android app plus an admin system to run the operational side, built fast enough to launch within a few weeks, without cutting corners on real-time order handling or payments. On the customer side, I built OTP-based authentication, food browsing with categories and search, cart management, and a full Razorpay integration covering UPI, cards, net banking, and wallet payments. I also added an in-app wallet system with real-time balance tracking, so users could pay directly from wallet balance instead of re-entering payment details every order. The bigger challenge was the admin panel, it needed to feel real-time, not just functional. Orders had to appear instantly, move through Received → Preparing → Out for Delivery → Delivered without lag, and give the business owner a clear picture of what was happening at any moment. I built this with live order management, food/menu control, and an analytics dashboard covering daily sales, revenue, and best-selling items, so the client could make decisions without digging through raw data. Keeping the two apps in sync, customer actions reflecting instantly on the admin side, required careful backend design around order state and notifications. The result was a production-ready MVP: secure payments, live order tracking, and an admin system the client could operate from day one.
1
55
Cover image for NeuralNest began as a personal
NeuralNest began as a personal project , I wanted to see if an AI tutor could verify a student's understanding instead of just dumping explanations at them. Surfboard Ventures picked it up, looking for a ready-to-use AI study tool rather than starting from zero. The hardest problem was making the tutor genuinely teach, not just answer. I built a LangGraph workflow where a grading node checks comprehension after every explanation, classifies the response as understood, confused, partial, or a doubt, and routes the student accordingly, a simpler analogy, a deeper breakdown, or a doubt-handling node, without losing state. Getting these classifications reliable took real prompt tuning. Retrieval was another challenge. Students upload PDFs, DOCX files, and images (OCR'd via GPT-4o Vision), and the tutor pulls context from the right scope, one session, selected materials, or the entire knowledge base. I used a multi-namespace Pinecone setup with Cohere embeddings to keep data isolated while enabling flexible retrieval. For exam prep, a pipeline parses past-year papers, ranks topics by frequency, and prioritizes high-yield topics in the study plan, recalibrating if quiz scores slip. The result: a full agentic study platform, topic extraction, a visual roadmap, adaptive quizzes, mastery scoring, and full tracing, that Surfboard Ventures was happy to take over.
1
74