GeezerVison

Hi, I'm Susant Achary 👋 📍 India ↔ Building Globally | 🤖 AI Systems Builder | 🚀 RecSys • GeoAI • Medical AI Tech Stack

Languages & App Development	Databases & Data Engineering
ML & Data Science	Domains & AI Systems

I build practical AI systems — from research and experiments to scalable pipelines, production APIs, and developer-facing applications.

Currently building in stealth_mode Port AI — secure, self-hosted, offline-ready AI across Mobile • Edge • Laptop • Raspberry Pi. Current Projects

🧠 Gemma Recipes - Practical Gemma notebooks and workflows for real-world AI use cases.
🎯 Video Point Tracker - Local-first multimodal video point tracking with React, Express, Docker, and support for Ollama, LM Studio, and llama.cpp.
👗 StyleSense - Fashion recommendation system covering real e-commerce personalization scenarios.
🛰️ Machine Learning for Geospatial - GeoAI learning resources, experiments, and spatial machine learning references.
🤗 HuggingFace HandsOn Cookbook - Hands-on Hugging Face examples across text, vision, speech, and multimodal AI.
⚡ MLX-One - Native multimodal inference on Apple Silicon with MLX, bringing text, vision, audio, and embeddings into a unified Python library and API.
📚 LLM Resources, Papers, Frameworks & Tools - Curated hub of LLM papers, tools, frameworks, and implementation resources.

Hugging Face Contributor

- Contributing to Hugging Face ecosystem (docs, tooling, model workflows) - Building reproducible AI demos + practical ML systems for developers 🚧 Building: Port AI (Secure • Self-hosted • Offline)

Port AI is my ongoing project to run Secure, self-hosted AI across Mobile • Edge • Laptop • Raspberry Pi: - 🔒 Secure-by-default: local-first, privacy-first, offline-ready - 🧠 LLM Apps: chat, RAG, tool-use/agents, eval harness - ⚡ Optimized: quantized models, streaming, caching, lightweight deployments - 🧰 Developer-friendly: FastAPI services, Docker profiles, reproducible configs Repo updates: architecture, demos, benchmarks, and deployment recipes.

🔍 What I work on - Large Scale Recommendation Systems & NLP - Computer Vision & Deep Learning - Edge + Mobile 🎯 Focus Areas 🩺 Medical AI

- Clinical text & imaging workflows - Decision support, triage signals, outcome prediction 🎯 Recommendation Systems

- Retrieval → ranking → evaluation (lift, deciles, offline/online metrics) - Cold-start, personalization, explainability 🌍 GeoAI / Geospatial

- Satellite imagery analytics & spatial ML - Location intelligence and geospatial pipelines 🔍 What I work on - Large-scale Recommendation Systems & NLP (ranking, retrieval, personalization) - Computer Vision & Deep Learning (feature learning, classification, detection) - Edge + Mobile (Android occasionally, on-device/optimized inference) - Applied AI domains: Healthcare • Education • E-Commerce • Agriculture • Supply Chain 🧰 Tech stack Core: Python • SQL • Bash ML: NumPy • scikit-learn • Matplotlib • (Deep Learning: Keras / PyTorch) Data: PostgreSQL Cloud: AWS • Azure Tools: VS Code • Git

📌 Focus areas - 🩺 Medical AI: clinical text + imaging workflows, decision support, triage signals - 🎯 RecSys: candidate generation, learning-to-rank, cold-start, evaluation & lift - 🌍 GeoAI: geospatial ML, satellite imagery analytics, location-based intelligence ✍️ Writing & links - Writes on: - LinkedIn: https://www.linkedin.com/in/curious-susant/ - 📫 Email: [email protected] ⭐ Support & community If you find my work useful, consider starring a repo — it helps a lot ❤️

Support Me, Brew Good Vibes

Help Keep the Ideas Open Source

GeezerVison

Susant SSusantAchary

Achievements

Achievements

Pinned Loading

Uh oh!