x:0 y:0
Pittsburgh, PA
ML Engineer · Maker · Inventor
© 2025
MAKER
Available for the right problem

Pranav
Kanuparthi

I build things that don't exist yet — patented sensor platforms, multi-agent AI systems, new ways to think about hard problems. My mission is building AGI: intelligent systems that understand, act, and work alongside people.

Maker Agentic AI Industrial ML Edge AI Sensing Systems LangChain · LangGraph
5+
Patents Filed
$20M
Downtime Saved
98%
Pose Accuracy
CES
Innovation Award
Let's Build Something See the Work

The moonshot is the minimum bar, not the aspiration.

At my core, I'm a maker. I get a kick out of building things that didn't exist before — and I've been doing it since I was 21 when I co-founded a startup and started filing patents. The best version of any problem is the one where you have to invent the solution.

Nine years. One thread: making machines that genuinely understand their environment. I go from raw sensor hardware through signal processing, model architecture, and cloud deployment — without needing a handoff.

🏆 CES Innovation Award · YogiFi
🛠️
Builder First
Sensor platforms, patents, production AI — the title changes but the instinct doesn't: if it doesn't exist, build it.
🧭
Comfortable in the Fog
Fuzzy requirements, no precedent. That's where I do my best work. I've shipped products where the playbook didn't exist yet.
Idea → Demo, Fast
Multi-agent fault detection, neuromorphic architectures, adversarial defenses. Build first, then make it excellent.
🔬
First-Principles
When existing approaches don't cut it, I design new ones. Five patents aren't a hobby — they're a habit.
🧱
Full Stack
Raw sensor hardware → signal processing → model architecture → cloud deployment. No excuses, no handoffs.
🎯
Mission-Driven
Not chasing roles or titles. Working toward something specific: machines that genuinely understand their environment.

Where I've Built

TDK SensEI
Sep 2021 – Present
Pittsburgh, PA
Full-time · 3+ yrs
Formerly Qeexo
Senior Machine Learning Engineer
  • Architected a production-ready multi-agent AI system for industrial asset diagnostics: three-agent hierarchical architecture using AWS Bedrock, LangChain, and LangGraph for real-time fault detection in rotating equipment.
  • Eliminated approximately $20M in annual downtime losses for large manufacturing clients; improved data pipeline efficiency by 62% memory reduction and 20% latency improvement.
  • Drove cross-functional delivery across engineering, data, QA, and product; presented technical roadmap to executive leadership.
  • Mentored two engineering interns; delivered invited talk on embedded ML systems at ARM DevSummit 2024.
$20M Downtime Saved 62% Memory Reduction 20% Latency Improvement ARM DevSummit Speaker
Carnegie Mellon
May – Dec 2020
Pittsburgh, PA
Research · 8 mos
Graduate Research Assistant — Neuromorphic Computer Architecture Lab
  • Developed Temporal Neural Network (TNN) architectures for Keyword Spotting and time-series classification using PyTorch, targeting Neuromorphic Sensory Processing Units.
  • Achieved 73% accuracy on Spoken MNIST under strict neuromorphic power constraints.
Advisor: Prof. John Shen 73% Spoken MNIST Accuracy
Wellnesys Inc.
Oct 2018 – Jan 2020
Bangalore, India
Co-Founder · CIO
Co-Founder & Chief Innovation Officer
  • Built YogiFi, a CES Innovation Award-winning patented AI and pressure sensor platform for real-time yoga pose classification and personalised guidance. Led a twelve-member cross-functional team.
  • Engineered CNN on pressure matrix data achieving 98% pose classification accuracy; technology underpins 3 granted patents.
CES Innovation Award 98% Pose Accuracy 3 Patents 12-person Team

Things I've Built


Patents & Publications

5
Patents Filed

3 Granted
2 Pending
Granted
System and non-intrusive method for exercise posture detection
US 20200254299 A1 · IN 201941005350
Granted
System and method to monitor an exercise posture of at least one user
IN 201841006784 · WO 2019162743A1
Granted
A flexible substrate and a method thereof
IN 201841006773 · WO 2019162742A1
Pending
System and method for performance evaluation of an exercise performed by a user
IN 202041000701
Pending
Real-time offside offense determination system and a method thereof
IN 201841046250
Talk
Embedded ML for Industrial IoT Applications — ARM DevSummit 2024
Paper
The Changing Nature of Human Adjudication
ITSMTGPD 2019 · IEEE Technology and Society Magazine

The Maker Doesn't Switch Off

The same instinct that makes me build software also puts me on a sailboat or at a shooting range. A maker doesn't switch off — they just find new problems to solve.

On the Water
US Sailing Certified · Sailor

Reading wind, water, and weather in real time. There's no undo button — same as a hard engineering problem. You work with what the environment gives you.

01
🏎️
Motorsport
Speed & Engineering

Machines pushed to their limits. Half the fascination is the engineering underneath — systems optimized to the edge of what's physically possible.

02
🎯
Sporting Clays
Focus & Timing

Every target is a slightly different problem. You can't fake preparation and you can't think too much. Total presence, every time.

03
🥊
Boxing
Discipline & Craft

Footwork, timing, controlled aggression under pressure. The gym teaches you that preparation isn't optional — it's the whole game.

04
I make things.
In the lab, on the water,
in the ring, everywhere.
The same instinct. Different problems. Always building.

Skills

Knowledge Areas
  • Deep Learning
  • Machine Learning
  • Agentic AI Systems
  • Sensor Systems Dev.
  • Data Analytics
Frameworks
  • PyTorch · TensorFlow
  • LangChain · LangGraph
  • PySpark · Pandas
  • Scikit-Learn · Pydantic
  • NumPy · Keras
Cloud & Infra
  • AWS Bedrock
  • Lambda · S3 · DynamoDB
  • ECR · Docker
Languages
  • Python
  • C / C++
  • SQL

Let's build
something 10×.

Looking for senior ML and agentic AI roles where the problem is hard, the stakes are real, and the solution doesn't exist yet. If that's what you're building — let's talk.

Send a Message