Samuel Raj is an AI Fullstack Engineer at HS Global AI, based in Hyderabad, building enterprise AI products spanning on-device inference, real-time voice pipelines, secure licensing infrastructure, multi-tenant backends, and generative avatar pipelines.

His work includes building digital human software: an offline digital human platform deployed to retail and healthcare clients across India, with 29 languages running entirely on device and no cloud speech path in the code, three-tier intent routing grounded in catalog-derived vocabulary, and cryptographic device licensing with Secure Enclave time attestation.

Key systems built by Samuel Raj include in-process LLM inference on Apple MLX and Metal, real-time WebRTC voice pipelines via LiveKit, hybrid retrieval fusing vector embeddings and full-text search by reciprocal rank fusion inside Postgres, entity and relationship extraction into a bi-temporal knowledge graph, secure JWT licensing with hardware attestation, multi-tenant SaaS backends with row-level security and RBAC, and ComfyUI avatar generation pipelines on Wan 2.2 Animate, LTX 2.2 and FLUX Klein.

Core technologies: Python, Rust, TypeScript, Apple MLX, NVIDIA CUDA, ComfyUI, Tauri, WebRTC, LiveKit, Whisper, AI4Bharat IndicConformer, Chatterbox TTS, pgvector, FAISS, Azure (Key Vault, Service Bus, GPU VMSS), Terraform, FastAPI, Next.js and Supabase.