ROBOTICS SOFTWARE · AUTONOMY · APPLIED AI

I build software for machines that have to make decisions in the real world.

M.Tech Automation & Robotics at DIAT Pune. I work across robot navigation, safe reinforcement learning, computer vision and edge inference—with an engineering bias toward measurable behavior, failure analysis and reproducible systems.

PRIMARYRobotics + AI Software
LOCATIONPune, India
30-second recruiter overviewArchitecture explainer only — not fabricated live-system footage.
01 / SELECTED WORK

Only the work that supports the roles I want next.

M.TECH RESEARCH · ACTIVE

Safe RL navigation for autonomous mobile robots

A research system combining Soft Actor-Critic with a Control Barrier Function safety filter in a warehouse navigation environment with static and dynamic obstacles.

201-dimensional observation: robot state, goal context and 192-beam LiDAR.
Continuous linear/angular control with an explicit CBF-QP intervention layer.
HER replay, trajectory diagnostics, safety-intervention logging and ablation-driven evaluation.
Research status is active; evaluation and baseline comparison are ongoing.
PythonPyTorchSACCBF-QPHERLiDAR
Watch the 36-second architecture walkthroughExplainer based on the project architecture. Research is active; no fabricated success claims.
DECISION PIPELINE / SAFE AMR
Observation201 dims
SAC policynominal action
CBF-QPsafety filter
Robot controlv, ω
COMPUTER VISION · EDGE AI

EdgeVision AI

A real-time computer-vision inspection interface built around live detection, system telemetry, alerting and deployment-oriented model benchmarking.

Live detection pipeline with a recruiter-readable operational dashboard.
ONNX Runtime deployment path with FP32 vs INT8 benchmark emphasis.
FastAPI + WebSockets architecture for live inference and system telemetry.
YOLOv8ONNX RuntimeOpenCVFastAPIReactWebSockets
Watch the 34-second system walkthroughSystem explainer based on the project architecture, not captured production footage.
EDGE INSPECTION / LIVE VIEW
INFERENCEONNX
MODEINT8
STREAMLIVE
ROBOTICS SYSTEMS

Control, kinematics and robot-software workbench

Selected systems work spanning ROS 2 services/actions, unicycle control, LQR, manipulator dynamics, impedance control and simulation-oriented robotics software.

ROS 2 communication patterns and robot-software fundamentals.
Control work across mobile robots, manipulators and dynamic systems.
C++ for robotics/DSA systems work; Python for AI, perception and research.
ROS 2C++PythonLQRKinematicsControl
SYSTEMS / ROBOTICS SOFTWARE
ROS 2services · actions · interfaces
ControlLQR · PID · impedance
Kinematicsframes · Jacobians · dynamics
Simulationnavigation · perception · testing
02 / ENGINEERING DEPTH

The details recruiters can ask me about.

01

Why SAC?

Continuous actions, off-policy sample reuse and entropy-regularized exploration for robot control.

02

Why a CBF layer?

To separate nominal learned behavior from an explicit safety constraint and log interventions.

03

Why edge deployment?

Latency, runtime efficiency and model quantization matter when perception has to run near the machine.

03 / ROLE FIT

Where this portfolio is aimed.

01
Robotics Software EngineerROS 2 · C++/Python · navigation · control · simulation
02
Autonomy / Perception EngineerLiDAR · computer vision · state/goal observations · real-time decision systems
03
Applied AI / Edge Vision EngineerYOLO · ONNX · FastAPI · model evaluation · deployment
CONTACT

Let’s talk about systems that move, see or decide.

I’m looking for engineering opportunities where robotics and AI are treated as systems problems—not just model demos.

Email me