Chitvan
Patel

I build |

Building portable MRI technology at neuro42. Firmware, APIs, image processing, cloud infrastructure. Also building and improving agentic AI workflows. Based in San Francisco.

View Projects
3+ Years Experience
6+ Projects Shipped
16+ Years of Tennis
About

Engineer at neuro42, working at the intersection of hardware and software.

Chitvan Patel at Golden Gate Bridge, San Francisco

I am currently an Embedded Software Engineer at neuro42, where I help build software for portable MRI devices. Previously, I was at Tarana Wireless and did LLM research at Santa Clara University. I did my undergrad in Electrical Engineering in Gujarat, India, then got my Master's in ECE from Santa Clara University.

Outside of work, I'm building agentic AI workflows, playing tennis (16+ years and counting), and gaming way too much Counter-Strike, PUBG, and Sims.

Here are some technologies I have been working with:

Python C/C++ Embedded Systems Medical Devices AI/ML Agentic AI
Experience

Embedded Software Engineer

neuro42, Inc.

Feb 2024 - Present
PythonC/C++JenkinsREST APIHIPAA

Hardware Integration Test Engineer

Tarana Wireless, Inc.

Nov 2023 - Feb 2024
PythonFPGAsPCBATE

Research Assistant, LLM

Santa Clara University

Sep 2023 - Nov 2024
PythonRAGElastic SearchNLP

Design Validation & Test Intern

Tag-N-Trac

Jun 2022 - Dec 2022
PythonTableauSerial
Projects
LLM Evals

gradetrail

Distributed LLM evaluation harness with content-addressed response caching, versioned judges, and interchangeable local/Ray execution backends. Re-scoring after a scorer fix costs $0.00 and takes under a second.

PythonRayAsyncioSQLiteAnthropic API
Featured

Brain-Computer Interface

Emotion recognition from EEG brain signals using CNN and LSTM. FFT for feature extraction. LSTM gave best accuracy across train-test splits.

PythonCNNLSTMEEGFFT
Embedded

Grill Thermometer

G070 board with meat probe thermocouple, EEPROM calibration, UI for target temps by meat type and doneness.

CAssemblyG070EEPROM
Computer Vision

Real-time Age, Gender & Emotion Detection

Haar Cascades for face detection, Caffe models for age/gender, custom model for emotion classification.

PythonOpenCVCaffe
Embedded

GPS Embedded System

ARM Cortex with RS485 data streams, fiber optic links to remote display modules, GPS coordinate reporting.

CARM CortexRS485GPS
Verification

8x8 Internet Switch Verification

UVM testbench with full functional coverage across SMOKE, 1-PORT, 4-PORT, and CONCURRENT tests.

SystemVerilogUVMSynopsys VCS
NLP Private

RAG Teaching Assistant

Chatbot for students using Retrieval Augmented Generation with Elastic Search and optimized vector embeddings.

PythonRAGElastic SearchLLMs
Agentic AI In Progress

Agentic Workflow Engine

A 3-layer architecture separating directives, orchestration, and deterministic execution. LLMs handle decision-making while Python scripts handle reliable execution. Self-annealing error loops that fix scripts, update directives, and get stronger over time.

PythonLLMsClaudePrompt EngineeringAutomation
Skills & Tools

Languages

PythonC/C++C#VerilogJavaScriptBash

Tools & Platforms

GitJenkinsDockerKubernetesAWSGCPLinuxJira

Frameworks

ReactREST APIs.NETElastic SearchCI/CD

Protocols

I2CSPIUARTUSBRS485RS232ADC/DAC

Domains

Embedded SystemsMedical DevicesRTOSHIPAAISO 13485IEC 62304

AI / ML

Deep LearningNLPComputer VisionRAGLLMsAgentic WorkflowsPrompt Engineering
Writing
July 8, 2026 · 9 min read

A 1-in-5 flake: debugging a SQLite WAL race under Ray

A test failed one time in five. The obvious fix, suggested by Claude, was wrong. Here's how a 30-line reproduction script found the real cause.
July 8, 2026 · 11 min read

Building Portable MRI: Where Software Meets Physics

The challenges of writing software for medical imaging devices, from real-time constraints to regulatory compliance.
Coming Soon

RAG Systems in Production: Lessons Learned

Building a RAG chatbot from scratch, vector embedding optimization, and real-world retrieval challenges.

Coming Soon

Agentic Workflows: Making LLMs Actually Reliable

Why putting LLMs in charge of everything fails, and how a 3-layer architecture with deterministic execution fixes it.

Get In Touch

Open to opportunities

Looking for my next role in embedded systems, medical devices, or forward deployed engineering. Let's connect.

GitHub LinkedIn Email