About Skills Experience Projects Education Contact Hire Me →
Open to opportunities · Bengaluru, India

Jesal
Patel.

Systems & Performance Engineering  ·  System Integration & Validation

Systems engineer, most recently at Qualcomm on the Cloud AI accelerator line (AI100, AI100U, AI200) — reading fleet telemetry, tracing faults through the compiler and runtime stack, and turning the findings into tooling and release criteria adopted across 5+ teams. Roots in EV motor-control firmware and hardware; now building local-first products like KutumbTree and LetsGo. Comfortable from silicon to application layer.

AI / ML Embedded Systems LLM Evaluation Generative AI Python
85%Fewer crashes
260+Tests automated
23h→12hPipeline runtime
5+Teams enabled
Jesal Patel
01 /  About Me

What drives me

"

दु:खेष्वनुद्विग्नमना: सुखेषु विगतस्पृह: | वीतरागभयक्रोध: स्थितधीर्मुनिरुच्यते

Bhagavad Gita 2.56 — One who is undisturbed by misery, unelated by happiness, free from attachment, fear, and anger — that one is a sage of steady mind. This verse guides how I approach both software and life.

AI / ML Engineering

Tuning and benchmarking CV, NLP, machine translation and segmentation models on Qualcomm Cloud AI accelerators, and writing the per-precision release gates (FP16, INT8, MX6).

Product Engineering

Founder of KutumbTree (family graph with an MCP server for LLM clients) and LetsGo (local-first, encrypted, CRDT-synced) — both shipped end to end.

Automation at Scale

Automated 260+ of ~300 test cases in Python and Jenkins, cut a 300-configuration suite to 105, and reduced manual validation by roughly 98%.

Embedded Systems

Deep roots in bare-metal firmware, motor controllers, and hardware validation across ARM, DSP, and TI-C2000 architectures.

02 /  Expertise

Skills & Technologies

A broad stack spanning AI/ML research, embedded firmware, DevOps, and hardware protocols.

Languages

PythonShell / Bash GroovyC

AI / ML

PyTorchTensorFlow ONNXRTCUDA TensorRTHugging Face LangChainRAG Agentic AIOpenCV Letta APIs

OS & Infra

Ubuntu 22.04RHEL 8.6/9.0 CentOS 7AWS AL2 KVMBareMetalWindows

Processor Arch

ARM Cortex M4/M0BCM2835 DSP / NSPTI-TMS320C2000

Protocols

SPII²CUART CANPWMRS485 WiFi 802.11Bluetooth 6.0

Tools & DevOps

JenkinsDocker GitJIRA PowerBIMatlab KiCADLTSpice Code Composer
03 /  Career

Professional Experience

From characterizing AI accelerators at fleet scale to shipping local-first products end to end.

Independent Products · Bengaluru Apr 2026 – Present
Founder & Lead Engineer — KutumbTree · LetsGo
  • Two shipped collaborative data systems, both centred on multi-party sync and merge.
  • KutumbTree, agent interface: built and published a Model Context Protocol server exposing the family graph as typed tools to LLM clients — schema design, tool contracts, and integration tests against live client calls.
  • KutumbTree, shipped and hardened: Node/Express and SQLite with a custom SVG renderer (collision-free layout, branch collapse, search-to-zoom) at 185+ nodes; live on web, Android build in Play Store testing. Security audit fixed a hardcoded JWT secret fallback, an unauthenticated webhook, and wildcard CORS; added security headers, rate limiting, and output escaping.
  • LetsGo, local-first architecture: no backend, no telemetry, a WebCrypto-encrypted vault; opt-in, user-owned sync via self-hosted WebDAV or a team backend with four-tier roles and live Loro CRDT sync. React and TypeScript in an Electron shell.
Qualcomm India Pvt. Ltd. · Bengaluru May 2022 – Feb 2026
Senior Software Engineer — AI Accelerator Systems
  • Used fine-grained telemetry (DSP utilization, DDR spillage, CPU affinity) to characterize accelerator behaviour at fleet scale; led root-cause work that cut crash incidents by 85% — the triage steps became the team playbook.
  • Tracked precision regressions through the compiler and runtime stack with layer-wise bisection, and filed accuracy-critical findings back to the compiler team.
  • Restructured the characterization pipeline, bringing a full run down from 23 hours to 12.
  • Built a performance-metrics framework (accuracy, latency, power, DDR bandwidth) that replaced legacy tooling and was adopted by 5+ teams.
  • Wrote the per-precision acceptance thresholds that became the release gate for the accelerator line (FP16 99.9%, INT8 99%, MX6 97% of FP32 reference).
  • Owned the Python + Jenkins nightly automation: automated 260+ of ~300 test cases and cut a 300-configuration suite to 105 without losing coverage — roughly 98% less manual validation.
  • Tuned and benchmarked CV, NLP, machine translation and segmentation models, from HVX/HMX tuning and DDR-spillage reduction up to day-to-day Python tooling.
  • Held system integration test sign-off for releases; led five engineers and mentored two, one promoted within a year.
Skandysys Pvt. Ltd. · Bengaluru Aug 2020 – May 2022
Software Engineer — Test Automation
  • Built a modular Python automation framework for ADAS embedded software, cutting time to market by 30% and manual testing effort by half.
  • Raised test coverage by 40% through automated suites, and wired the framework into CI/CD to halve manual intervention.
GreenVolt Mobility · Ahmedabad Dec 2016 – Jun 2020
Development Engineer → VP Electronic Systems → President
  • Led hardware and firmware for EV motor-control systems — PMSM controllers, custom test instruments, and pre-production and post-assembly drive diagnostics.
  • Programmed control and diagnostic tooling in C and Python against TI Control Suite and Code Composer Studio; built Python validation tooling that halved manual test effort.
  • Designed schematics and PCBs in KiCad and Eagle, and ran dyno testing and electrical parameter analysis to check operating consistency.
04 /  Portfolio

Projects & Highlights

A selection of systems built, frameworks shipped, and problems solved across AI, hardware, and software.

May 2026 · WIP
KutumbTree
Personal Project

A web app to build, visualize, and explore your family tree — making it easy to map generational connections and preserve family history digitally.

Full StackWeb AppLive ↗
2025 · Paused
PlantPause — Mindful Gardening Brand
Personal Project

An early-stage brand and product for urban gardeners, centred on "Grow slow, grow mindful." Built end-to-end — web presence, content blog, and community channel — with full ownership of product direction and engineering. The project is currently on hold.

Full StackWebProduct
Mar 2024
AI Model Accuracy Validation Framework
Qualcomm India

A fully automated pipeline that validates accuracy for CV, NLP, machine translation and segmentation workloads on Qualcomm AI100/AI100U/AI200 hardware, against GPU-generated FP32 golden references. What once took days now runs in hours — end-to-end, from model input to accuracy report, via a single CLI command.

PythonJenkinsLLMAI100
Nov 2023
Deep Learning Profiling Framework
Qualcomm India

A modular benchmarking tool that measures latency, accuracy, DDR bandwidth, and power for AI models running on AI100. Replaced ad-hoc scripts with a structured, shareable system — reducing profiling time by 65% and enabling cross-team reuse.

PyTorchONNXRTBenchmarking
Sep 2019
PMSM Motor Controller Validation
GreenVolt Mobility

Automated EV motor controller tests using a Raspberry Pi-based dyno rig. Cut manual testing time by 50% and improved defect detection by 30%. The same engineering discipline I'd later apply at scale at Qualcomm.

CRPiTMS320KiCAD
Apr 2017
3D Printer — Final Year Project
Indus University

Designed and assembled a functional 3D printer from scratch — stepper control, RAMPS 1.4 board, G-code firmware, and custom hardware mounts. A first-principles project that sparked a lifelong love for building things that work.

StepperRAMPS 1.4G-code
Mar 2017
ABU Robocon — Autonomous Robot
MIT Academy, Pune — National Competition

Represented college at a national robotics competition. Designed an autonomous navigation system with real-time motor control and sensor fusion. Competing on a national stage as an undergrad was a defining experience in building under pressure.

Embedded CMotor ControlRobotics
05 /  Learning

Education

Always learning — from embedded systems to generative AI, and everything in between.

Completed
Executive Post Graduate Programme — Applied AI & Agentic AI
IIIT-Bangalore & upGrad
Overall score: 95.77%  ·  Feb – Aug 2026
Focus: LLM Architecture · Agentic Systems · RAG · AI Product Development
Completed
PG Diploma — Machine Learning & Artificial Intelligence
IIIT-Bangalore & upGrad
GPA: 3.45 / 4.00
Coursework: Deep Learning · AI Systems · Model Deployment
Completed
B.E. — Electronics & Communication Engineering
Indus University, Ahmedabad
GPA: 7.9 / 10
Coursework: Signal Processing · Microcontrollers · FPGA Design
Completed
Diploma — Electronics & Communication Engineering
VPMP Polytechnic (LDRP), Gandhinagar
GPA: 7.0 / 10
Coursework: Embedded Systems · VHDL · Communication Protocols
06 /  Content

On YouTube

I share what I build and learn — from AI projects to Python tutorials — in plain, practical terms.

rubixcodelab
@rubix-coder

A channel about building real things with code — AI/ML projects, Python deep dives, automation ideas, and hands-on tutorials. The kind of content I wished existed when I was learning: practical, direct, and honest about what actually works. From writing your first Python script to fine-tuning language models and building agentic AI systems.

Visit Channel