buddywhitman

about
resume

Silicon to
Intelligence.

Pulkit Kumar

FULL-STACK INFERENCE ARCHITECT

Architecting production-grade infrastructure from **RTL-level hardware** up to **agentic AI orchestration**. Specializing in low-latency systems for HFT, hospital networks, and autonomous racing. Q1 Nature Author.

Pulkit Kumar - Full-Stack Inference Architect

2026 / Performance Architecture

I architect production-grade infrastructure where **determinism meets intelligence**. My work spans from **RTL-level hardware optimization** for low-latency controllers to **agentic AI orchestration** for safety-critical environments.

Currently, I'm a founding engineer at **Fettle**, where I lead software architecture for distributed voice AI. I also lead automotive electronics for **SolarMobil**, designing stochastic telemetry engines that fuse vehicle state with environmental data in real-time.

Strategic Engineering

Solar Racing Telemetry & Race Engineering Digital Twin

Solar Racing Digital Twin

Engineered a high-performance race strategy engine for solar EVs. Built a real-time digital twin integrating weather APIs, GIS elevation data, and vehicle dynamics. Leveraged stochastic receding horizon control for SoC optimization, leading to IEEE VTC publication and 2+ provisional patents.

HPCControl TheorySTM32H7RTOSC++InfluxDB
Agentic Voice AI for Hospital Workflows

Fettle: Agentic Voice AI

Architected distributed voice AI agents for high-concurrency hospital customer experience. Implemented SIP trunking with LiveKit and optimized real-time inference using PyTorch & TorchRec. Founding Engineer owning the entire DevOps lifecycle and backend architecture.

Agentic AILiveKitPyTorchKubernetesFastAPIInference
CoreEL SoC Design & ASIC Flow

SoC Design & ASIC Flow

RTL2GDSII design of high-speed CAN bus and I2C controllers. Collaborated with synthesis and UVM verification teams to ensure timing closure on a Virtex 7 FPGA. Expert exposure to ASIC design flow including STA, floorplanning, and PPA optimization.

VerilogFPGAUVMASIC FlowVivadoHFT
Multimodal Biofeedback System (Nature Publication)

Biofeedback Edge AI

Technical validation of a multimodal emotion-adaptive biofeedback system for autonomic regulation. Published in Nature Scientific Reports (Q1). Deployed multimodal edge AI pipeline using LSTM + transformer encoders on Jetson/ESP32 platforms.

Edge AIDSPNatureTransformerJetsonC

Architecting the High-Stakes.

Actively seeking roles in **HFT**, **Inference Infrastructure**, and **Safety-Critical Embedded Systems**. I build the systems that cannot fail.