Akash Kandi | AI/ML Engineer

Robotics Software Engineer

AKASH
KANDI

Currently @ Qualcomm · M.S. CS — University of Georgia

Building autonomous robotics software for perception, navigation, motion planning, and embedded systems — from edge AI to digital twin simulation.

Get in Touch

3+

Years Experience

3

Live Projects

28%

Nav Accuracy Gain

Scroll to explore

About Me

WHO
I AM

I'm Akash Kandi, a Robotics Software Engineer with 3+ years of experience building autonomous systems for perception, navigation, motion planning, and embedded control.

At Qualcomm, I engineer ROS 2-based navigation and real-time perception pipelines with SLAM, YOLO, and sensor fusion — improving navigation accuracy by 28% and cutting inference latency by 37%. Previously at KPIT Technologies, I built autonomous mobility software for cameras, LiDAR, IMU, and GPS sensor fusion.

M.S. Computer Science, University of Georgia (GPA 3.71). Open to relocate. Passionate about edge AI, digital twins, and real-time robotic control.

C++PythonROS 2 SLAMMoveIt 2OpenCV CUDADockerKubernetes

Technical Skills

WHAT
I KNOW

Programming

C++ · Python · CUDA · MATLAB

Robotics Frameworks

ROS 2 · Nav2 · MoveIt 2 · ROS 2 Control · Micro-ROS · DDS / Fast DDS · BehaviorTree.CPP

Autonomy & Planning

SLAM · Visual SLAM · Localization · MPC · RRT* · CHOMP · OMPL · Behavior Trees · Multi-Robot Coordination

Perception & AI

OpenCV · PyTorch · TensorFlow · YOLO · Sensor Fusion · 3D Mapping · Visual-Inertial Odometry · Reinforcement Learning

Simulation

Gazebo · Isaac Sim · Isaac Lab · NVIDIA Omniverse · Digital Twins · HIL / SIL Testing

Embedded & Control

FreeRTOS · ARM Cortex · FPGA · Motor Control · CAN Bus · EtherCAT · Embedded Linux

Infrastructure & Deployment

Docker · Kubernetes · Git · GitHub Actions · GitLab CI/CD · ONNX Runtime · TensorRT · Edge AI Deployment

Experience — 1 of 2

QUALCOMM
USA · JAN 2026 – PRESENT

Robotics Software Engineer

January 2026 – Present
United States

ROS 2SLAMNav2 MoveIt 2CUDATensorRTKubernetes

ROS 2 autonomous navigation with Nav2, BehaviorTree.CPP & SLAM — navigation accuracy +28%, mission time -22%
Real-time perception pipelines using C++, OpenCV & YOLO on LiDAR + multi-camera data — detection precision +31%, latency -37%
MoveIt 2, OMPL, CHOMP & MPC motion planning — trajectory planning -35%, manipulation success +24%
Embedded Linux robotics with Micro-ROS, FreeRTOS & EtherCAT — control-loop latency -33%, reliability 99.8%
Digital twin & HIL/SIL validation in Gazebo, Isaac Sim & NVIDIA Omniverse — field validation effort -41%
Kubernetes + TensorRT + CI/CD deployment — release cycles +45% faster, deployment failures -32%

Experience — 2 of 2

KPIT
TECHNOLOGIES

Robotics & Perception Engineer

Jan 2022 – Jul 2024
India

C++Visual SLAMOpenCV TensorRTDDSGazeboDocker

Perception software for autonomous mobility using C++, OpenCV & Visual SLAM — environmental accuracy +27%
Multi-sensor fusion of camera, LiDAR, IMU & GPS — vehicle pose estimation error -32%
Autonomous path planning with RRT*, Localization & Behavior Trees — route replanning time -29%
High-performance inference with TensorFlow, CUDA & TensorRT — perception execution +43% on embedded platforms
Gazebo & Digital Twin validation environments — test coverage +46%, on-road validation effort -35%
GitLab CI/CD, Docker & Git automation — build/validation cycles -38%

Projects — 1 of 2

WHAT
I BUILT

StockMind

Multi-Agent Investment Research System

4 parallel agents — news, SEC filings, financials & FinBERT sentiment — synthesize a full investment report with BUY/HOLD/SELL in under 60 seconds. Live WebSocket dashboard streams each agent's status in real time.

LangGraphFastAPIReact 18FinBERTPostgreSQLWebSocketsDockerGPT-4o-mini

Live Demo ↗ GitHub API Docs

WatchLess

YouTube Channel RAG Knowledge Base

Ingest any YouTube channel → semantic Q&A with timestamp citations across 2,000+ chunks. Unique Cross-Video Synthesis detects when answers span multiple videos and explains how each approaches the topic differently.

LangChainFastAPIReact 18ChromaDBOpenAI EmbeddingsGPT-4o-miniDocker

Live Demo ↗ GitHub API Docs

Projects — 2 of 2

MORE
WORK

PersonaRank

Personalized Recommendation Engine

Collaborative filtering recommendation system using SVD matrix factorization, trained on the MovieLens 100K dataset (100,000 ratings, 943 users, 1,682 movies). Predicts user ratings for unseen items and returns personalized top-N recommendations via a FastAPI endpoint.

Test RMSE: 0.9337 · 80/20 train/test split

Pythonscikit-surpriseSVDpandasFastAPIDocker

GitHub ↗

Architecture

User ratings → SVD matrix factorization
→ latent factor decomposition
→ predicted ratings for unseen movies
→ top-N personalised recommendations
→ FastAPI /recommend/{user_id}
→ Docker containerised deployment

100K

Ratings trained

1,682

Movies indexed

0.9337

Test RMSE

Education

BACK
GROUND

M.S. Computer Science

University of Georgia · Athens, USA · May 2026

GPA: 3.71 / 4.0

Robotics · Deep Learning · Distributed Systems · Advanced Algorithms · Computer Vision · Embedded Systems

Contact

LET'S
CONNECT

Open to Robotics Software Engineer roles — autonomous mobility, perception, navigation, and embedded systems. Open to relocate.

linkedin.com/in/akash-kandi-23635841b github.com/akashkandi +1 (408) 800-5221 akashkandi582@gmail.com

📍 United States · Open to Relocate

GitHub LinkedIn Email