Students pick a base model by leaderboard rank and then discover the license forbids commercial use, or the model has no quantized GGUF build, or the community support is thin. Here's the checklist that actually matters.
- License: permissive enough for what you plan to do with the result, including a resume portfolio piece or a startup.
- Size vs your hardware: can you fine-tune it with QLoRA on the VRAM you actually have, not the VRAM a benchmark assumed.
- Instruction-tuned base, not raw pretrained, unless your task genuinely needs to start from scratch.
- Community tooling: an active ecosystem means faster answers when something breaks mid-project.
- Context length: matches what your actual inputs need, not the largest number on the spec sheet.
Leaderboard rank predicts general capability, not whether a model is fine-tunable on your setup. See fine-tuning on a consumer GPU: the VRAM budget and small language models: 1B–3B use cases.
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU Kanpur) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. Email pranjulrathour41@gmail.com.
