class ChandanKumar:
role = "AI Engineer | Voice AI"
focus = [
"Speech AI",
"LLMs",
"Real-Time Voice",
]
stack = [
"Python",
"Go",
"FastAPI",
"Ruby on Rails",
"LiveKit",
"WebRTC",
"SIP"
]
interests = [
"Conversational AI",
"Open Source AI",
"Noise Cancellation",
"Fine-tuning"
"Latency Optimization"
]
motto = "Ship models, not just notebooks"I build end-to-end ML systems, from research prototypes to production-grade pipelines. Whether it's fine-tuning a Mistral model, building retrieval-augmented generation systems, or deploying computer vision models for cell segmentation, I love turning complex AI problems into working software.
|
AI / ML
|
Backend & Infra
|
Data & Storage
|
A curated set of my pinned repositories — spanning LLM safety, speech, computer vision, medical AI, and quantization.
| Project | What it does | Stack | |
|---|---|---|---|
| ⚡ | Bitwiser | 1-bit quantization tooling for Apple Silicon | Quantization TypeScript |
| 🛡️ | ShieldPrompt | PII-masking middleware for LLMs (PyPI · MCP server) | LLM Safety PII |
| 👤 | Facematch | Face recognition & matching system | CV Recognition |
| 🎙️ | Vocat | Voice-based interview agent | Voice Multimodal |
| 🗣️ | whisper-finetune | LoRA fine-tuning of Whisper-small for Hindi ASR on Common Voice 17 | ASR LoRA PEFT |
Worked at ISRO → Built AI systems at annam.ai → contributed to turocrate.ai → then neurofin.ai
╔══════════════════════════════════════════════════════════╗
║ "The best model is the one that ships." — me, probably ║
╚══════════════════════════════════════════════════════════╝
Open to collaborations on AI/ML projects
