Building autonomous systems

Shifat Santo

Systems AI Engineer · Founder @ Rhemic AI

Building autonomous AI systems, LLM infrastructure, and applied AI products that operate at production scale. From low-level systems in C++ and Rust to full-stack AI applications.

// projects

Selected Work

Systems and products built from scratch, shipping in production.

Rhemic AI

Voice agent infrastructure and developer tools. Leading architecture, backend systems, and ML pipeline design as CTO.

TypeScriptPythonGo

VectorVault

HNSW vector search engine built from scratch in C++20. Custom memory allocator, SIMD-optimized distance calculations, lock-free concurrent access.

C++20HNSWSIMD

mcp-audit

Security auditing tool for Model Context Protocol servers. Static analysis of tool definitions and permission validation.

RustMCPSecurity

Soniq

Real-time voice agent platform with sub-200ms latency. Streaming STT, LLM reasoning, and TTS pipeline.

PythonWebSocketVoice AI

Bangla LLM

Training a large language model for Bangla. Custom tokenizer, web corpus pipeline, distributed training on cloud GPUs.

PythonPyTorchNLP

undertone

Audio sentiment analysis library published on PyPI. Spectral feature extraction and lightweight ML classification.

PythonPyPIML
// proof of work

By the Numbers

Shipped systems, production deployments, and measurable impact.

12+
Production Systems
50K+
Lines of Systems Code
200ms
Voice Pipeline Latency
60%
LLM Cost Reduction
Voice Agent Infrastructure
Building Rhemic AI's voice agent platform -- real-time STT, LLM reasoning, and TTS with sub-200ms end-to-end latency.
Custom Vector Database
HNSW-based similarity search engine in C++ with SIMD-optimized distance calculations and concurrent indexing.
Production Voice AI
End-to-end voice pipeline -- Whisper STT, custom VAD, real-time streaming -- deployed for call center analytics.
Full-Stack AI Products
Shipped end-to-end from low-level C++/Rust systems to Next.js frontends with LLM-powered features.
// about

Background

I'm a systems AI engineer and the founder of Rhemic AI. I study Computer Science at UT Dallas and spend most of my time building AI infrastructure and products from the ground up.

My work spans the full stack of AI systems — from custom vector databases in C++ and voice processing pipelines in Rust to production LLM applications and autonomous agent frameworks. I care about understanding every layer of the stack, not just using abstractions.

I believe the best AI products come from engineers who understand both the systems underneath and the humans on top. That's what I build toward.

Technical Stack

Languages
C++RustGoPythonTypeScript
AI / ML
LLMsRAGAgentsWhisperEmbeddingsFine-tuning
Infrastructure
DockerLinuxCUDAA100/H100FastAPI
Frontend
Next.jsReactTailwind
// contact

Get in Touch

Open to conversations about AI systems, infrastructure, and building things that matter.