Skip to Content
PK

PRANAV KOLEKAR

ROBOTICS · PERCEPTION · PHYSICAL AI

RésuméContact

/about

I build robots that see, plan and act

Perception, navigation and manipulation on ROS 2, validated sim-to-real in NVIDIA Isaac Sim, from robot-learning policies to the system architecture around them. Nav2 contributor.

now
Robotics Software Engineer · Neural Foundry
where
Hyderabad, India
open to
Robotics · Perception · AI robotics · Research · Solutions architecture

  • 6→15 Hz

    camera stream restored

  • 8

    devices in a firmware-in-the-loop twin

  • 50+

    demos trained into robot policies

01 · Systems & solutions

Integration and solution design

The contracts between firmware and the platform, and the recommendations that start before any code is written.

Embedded & systems

Contracts between firmware and the platform

  1. device firmware
  2. command + telemetry registry
  3. manifest
  4. MQTT
  5. automation engine

01

Command registry with preconditions

Commands declare the object states they need plus setup and cleanup hooks, so the automation engine no longer orchestrates them externally.

02

Telemetry registry

Devices declare the signals they report, with JSON schemas derived automatically from C++ types and published in the same manifest.

03

PLC architecture decision

Compared replacing per-device microcontrollers with a PC as central controller and a PLC distributing commands; chose PC → PLC over Modbus TCP and proved it in a proof of concept.

04

Lab instrument onboarding

Brought a pipette and an RS-232 weight balance onto the platform for a customer demonstration.

Solutions architecture

From a client's requirement to a recommendation

  1. requirement
  2. options
  3. trade-offs
  4. recommendation

01

Robot-cell specification

Specified arm class, vision, edge compute and safety for three industrial use cases, weighing ROS 2 driver availability, licensing, safety rating and import constraints.

02

Bill of materials for a statement of work

Selected the robot and edge device for a heavy-payload handling cell: one preferred option and two justified alternatives.

03

Instrument integration study

Assessed three integration paths for a laboratory analyser on licensing, warranty risk and architectural fit, ending in a written recommendation.

04

Site feasibility inspection

Inspected a production site for a cleaning-robot deployment and wrote the technical feasibility document.

02 · Case studies

Debugging case studies

Each one started as a symptom with no obvious cause. Here is where it sat in the pipeline and how it was fixed.

Simulation

The robot that flew off on spawn

  1. USD asset
  2. suction joint
  3. PhysX
  4. articulation

root cause · suction jointThe suction joint had no body1, so Isaac Sim pinned it to the world frame and baked a ~10 m world anchor into the asset.

fixShipped a workaround (suction joint deactivated); NVIDIA confirmed the defect and targeted a fix for 6.x.

~10 m

stale world anchor found in a tool joint

Read the case study →
  • NVIDIA Isaac Sim
  • OpenUSD
  • PhysX
  • Python

Perception

Camera stream under 6 Hz

  1. camera
  2. DDS QoS
  3. WebRTC
  4. operator

root cause · DDS QoSThe camera topic used Best Effort QoS, which drops samples whenever the system is under load.

fixSwitched the camera topic's QoS reliability policy to Reliable.

6→15 Hz

camera stream restored

Read the case study →
  • ROS 2
  • DDS QoS
  • WebRTC
  • NVIDIA Isaac Sim

Robotics · control

A gantry that missed its deadlines

  1. trajectory controller
  2. hardware interface
  3. firmware
  4. steppers

root cause · firmwareThe firmware did not finish each setpoint within one control period, so timing error accumulated; step-delay scaling made it worse.

fixMoved step timing into the microcontroller so every setpoint executes in exactly one control period, and dropped the scaling.

1/f

every setpoint in one control period

Read the case study →
  • ros2_control
  • Joint Trajectory Controller
  • C++
  • Arduino
  • Stepper motors

Navigation · open source

Phantom floor collisions in Nav2's simulation

  1. LiDAR sensor
  2. scan
  3. costmap
  4. planner

root cause · LiDAR sensorA simulated LiDAR misconfiguration turned max-range rays into floor hits.

fixCorrected the sensor configuration; the fix was reviewed and merged upstream.

#26

fixed and merged upstream

Read the case study →
  • Nav2
  • ROS 2
  • Gazebo
  • LiDAR

03 · Experience

Where the work happened

  1. Feb 2026 – Present · Hyderabad, India

    Robotics Software Engineer

    Neural Foundry

    Isaac Sim, ROS 2, mobile manipulation, and embedded C++ — building a cloud-based digital twin for lab-automation hardware.

    • ▸Architected a firmware-in-the-loop digital twin: a C++/Python Isaac Sim extension running real device firmware for an 8-device lab-automation cell, entirely in the cloud
    • ▸Built an end-to-end mobile-manipulation twin (differential-drive AMR + 6-DOF Doosan cobot) with full ROS 2 / ros2_control integration and vision-guided pick-transport-place; tuned Nav2 across 5 subsystems
    • ▸Root-caused an NVIDIA-confirmed Isaac Sim 5.1 defect through OpenUSD layer inspection and shipped the team workaround
    • ▸Benchmarked multi-instance Isaac Sim on L40S GPUs (120 → 8 fps at 8 instances) and isolated a ~4-core-per-instance CPU ceiling that set infrastructure sizing

    capabilities:

    case studies:The robot that flew off on spawnCamera stream under 6 HzA gantry that missed its deadlines

  2. Jun 2025 – Jan 2026 · Hyderabad, India

    Robotics Software Engineer

    Nxtwave Disruptive Technologies

    ROS 2, Autoware, manipulation, and imitation learning — extending an SDK for the Unitree Go2, building a manipulator control stack, and deploying Autoware on an F1TENTH vehicle.

    • ▸Extended the ROS 2 SDK for the Unitree Go2 with dual transport over WebRTC and CycloneDDS
    • ▸Built a LeRobot leader–follower teleoperation pipeline, collected 50+ demonstrations and trained ACT and Diffusion Policy models that ran repeatably on hardware
    • ▸Deployed the full Autoware stack on an F1TENTH vehicle: NDT localisation, lane-keeping and pure-pursuit tracking
    • ▸Implemented a serial control stack for a bus-servo 6-DOF arm with DH-parameter forward kinematics for real-time end-effector pose

    capabilities:

  3. Feb 2025 – Present · Finland (Remote)

    Robotics Developer

    Ondroid

    • ▸Developed Python/OpenCV automated computer vision pipelines for output verification, eliminating manual review and standardizing evaluation outcomes
    • ▸Integrated Pytest into the CI backend pipeline, increasing test coverage and enforcing regression-free deployments across multiple release cycles

  4. Feb 2024 – Jul 2024 · Pune, India

    Robotics Lead — Robocon

    M.E.S. Wadia College of Engineering

    • ▸Led hardware and software R&D for a two-robot team, integrating closed-loop motor control and custom control firmware on Robot 1
    • ▸Provisioned a complete ROS 2 navigation stack on Robot 2 — URDF modeling, wheel encoder odometry, 2D LiDAR, IMU integration, and Gazebo simulation for pre-competition testing
    • ▸Represented the team at ABU Robocon 2024, New Delhi

    capabilities:

  5. Mar 2024 – May 2024 · Pune, India

    Electronics Intern

    Canspirit AI

    • ▸Connected IoT sensor nodes to Arduino Cloud for real-time data acquisition
    • ▸Built cross-platform Python logging pipelines to streamline data storage, retrieval, and downstream analysis

  6. Oct 2023 – Jan 2024 · Pune, India

    Embedded Systems Developer

    Indkarta LLP

    • ▸Designed custom PCB layouts in KiCad for two embedded systems deployed in clinical medication-processing environments
    • ▸Owned firmware development from prototype through field testing

  7. Dec 2021 – May 2025 · First Class

    Bachelor of Engineering

    Electronics and Telecommunications · M.E.S. Wadia College of Engineering

04 · Projects & open source

Independent projects

Personal · 2024 – now

Terrain explorer

Tracked robot that explores GPS-denied spaces with a frontier-based strategy, fusing LiDAR and IMU in a custom ROS 2 pipeline to build real-time 3D occupancy maps.

View on GitHub →

Open source · merged

Nav2 simulation fix

Fixed phantom floor collisions caused by max-range LiDAR rays in the Nav2 minimal TurtleBot simulation.

Read the case study →

05 · Skills

Skills in practice

Each skill links to the chapter, role or case study where it was used.

Programming Languages

Core languages across robotics, embedded and tooling work.

Robotics Frameworks

Navigation, manipulation and autonomy stacks used in production.

Simulation

Digital-twin and debugging tooling for sim-to-real work.

Perception

Detection, pose estimation and multi-sensor fusion.

  • OpenCV

Robot Learning

Imitation learning shipped on physical hardware.

Systems & Infrastructure

Embedded contracts, cloud and delivery tooling around the robotics stack.

Hardware

Robots and interfaces worked with directly.

06 · Contact

Get in touch

Email

pranavkolekar13@gmail.com

0 characters (minimum 20)