name : Asad Ali
role : Frappe / ERPNext Solutions Engineer (Full-Stack)
company : Agrovisions, Faisalabad, Pakistan
education : BS Computer Science — FAST NUCES, CFD Campus (2021–2025)
expertise :
- End-to-end AI/ML systems: RAG pipelines, LLM fine-tuning, multimodal apps
- Full-stack mobile development with Flutter & Firebase
- Custom Frappe/ERPNext app development with AI integration
tagline : "Building intelligent systems — from LLMs to ERPs, one commit at a time."
open_to : [Collaborations, Hiring, Consulting, Mentoring]I engineer AI systems that ship — not just notebooks. From clinical RAG pipelines over MIMIC-IV clinical notes to a cross-platform creator economy app with a smart recommendation engine, I build things end-to-end and deploy them as live, interactive demos. On the ERP side, I've shipped 15+ production-grade Frappe/ERPNext custom apps, including an Invoice OCR pipeline that uses AI embeddings and fuzzy matching to map noisy scanned documents to ERP records.
Cross-Platform Creator Economy App 50+ trial users · 30% retention boost · 25% engagement lift via recommendation engine · Stripe payments integrated |
15+ Production Custom Apps Invoice OCR · Consignment Management · Commission Automation · AI embeddings + fuzzy matching |
Live AI/ML Demos All projects deployed as interactive Gradio / Streamlit demos · Documented on Medium |
AI / ML & NLP
Full-Stack Mobile & Web
ERP & Automation
Tools & Infrastructure
| Project | Description | Stack |
|---|---|---|
| 🎨 Visual Vault | Cross-platform creator economy app with smart recommendation engine — Final Year Project | Flutter · Firebase · Stripe |
| 🏥 DiReCT — Clinical RAG System | RAG pipeline over MIMIC-IV-Ext clinical notes using BM25 + FAISS retrieval and instruction-tuned LLMs (BioMistral / Nous Hermes) | Python · FAISS · BM25 · HuggingFace |
| 🧾 Invoice OCR (ERPNext) | AI pipeline using embeddings + fuzzy matching to map noisy scanned invoices to ERP items and parties | Frappe · Python · Embeddings |
| 📝 Roman Urdu Poetry Generator | LSTM-based poetry generation model deployed as a live Gradio demo | LSTM · PyTorch · Gradio |
| 🔁 Transformer Code & NMT Models | Custom Transformer for Pseudocode ↔ C++ translation and Arabic → English NMT | PyTorch · HuggingFace · Gradio |
| 🧩 AI Math Riddle & Meme Solver | GPT-2 + TinyLlama powered riddle and meme understanding app | GPT-2 · TinyLlama · Gradio |
| 🖼️ Ask-the-Image | Multimodal app combining Whisper (speech-to-text) + BLIP-2 (visual QA) + TTS output | Whisper · BLIP-2 · Gradio |
| 💼 AI Career Assistant | Voice-driven career advisor using Whisper + LLM + Gradio interface | Whisper · LLM · Gradio |
🔗 All projects: github.com/asadsandhu | 📝 Technical write-ups: medium.com/@asadsandhu
I document every major project with a deep-dive technical post — architecture decisions, implementation details, and lessons learned. No fluff.
| Topic Area | Blog |
|---|---|
| Clinical RAG Systems & LLM fine-tuning | @asadsandhu on Medium |
| Multimodal AI (Whisper, BLIP-2, TTS) | @asadsandhu on Medium |
| Transformer models — Code & NMT | @asadsandhu on Medium |
| Flutter & Firebase architecture | @asadsandhu on Medium |
| Frappe/ERPNext customization & AI integration | @asadsandhu on Medium |
| Degree | Institution | Year |
|---|---|---|
| BS Computer Science | FAST NUCES, CFD Campus — Faisalabad | 2021 – 2025 |
| Intermediate (Pre-Engineering) | Government Islamia Graduate College — Faisalabad | 2019 – 2021 |
- 🚀 Built Visual Vault — cross-platform creator app with 50+ trial users, 30% retention boost, and a smart recommendation engine driving 25% engagement lift
- 🏥 Engineered DiReCT — a clinical RAG system over the MIMIC-IV-Ext dataset using BM25 + FAISS + instruction-tuned LLMs; fully documented on GitHub and Medium
- 🏭 Shipped 15+ production Frappe/ERPNext apps including an AI-powered Invoice OCR pipeline using embeddings and fuzzy matching
- 🤗 All AI/ML projects deployed as live interactive demos on Hugging Face Spaces (Gradio / Streamlit)
- 🔬 Fine-tuned GPT-2, TinyLlama and built custom Transformer architectures for code translation and neural machine translation
I'm open to discussions and opportunities in:
- 🔭 RAG Systems & LLM Fine-tuning — architecture, evaluation, deployment
- 📱 Flutter & Firebase — production-grade mobile apps
- 🏭 Frappe / ERPNext — custom development and AI integration
- 🧠 NLP, Transformers & Multimodal AI — research to production
- 🚀 Hugging Face Spaces — shipping AI demos
Last updated: 2025 · github.com/asadsandhu