Prateeksha Ranjan

About

Robotics software engineer who built and owned core navigation components for a TurtleBot in ROS/ROS2 and Gazebo, including an RRT-based planner, waypoint-to-velocity motion planning, and PID trajectory tracking. Demonstrates strong real-time debugging skills (control-loop timing under CPU load), costmap/occupancy-grid tuning, and distributed ROS2 communication design using DDS/QoS, plus Docker and CI/CD automation experience from Keysight.

Experience

Education

Key Strengths

• Built end-to-end navigation stack components (RRT planner + waypoint-to-velocity motion planning) in ROS/Gazebo
• Deep debugging of mapping/costmap issues (inflation tuning, resolution mismatch) using visualization tools
• Improved trajectory quality via waypoint spacing control and curve-fitting smoothing
• Stabilized real-time robot behavior by analyzing control-loop timing under CPU load and enforcing deterministic callbacks
• Strong ROS2 integration skills (TF2 frames, nav/geometry messages, topics/services)
• Designed distributed/heterogeneous robot communication via ROS2 DDS interface layer and QoS tuning
• Reproducible robotics environments using Docker; experience with CI/CD automated validation (Keysight)
• Refactored Vue/TypeScript component library to standardize props/events and improve state management
• Improved reliability by isolating business logic from UI logic and adding unit test coverage
• Strong documentation practice with before/after examples, design notes, and layered docs for different audiences
• Systematic issue diagnosis: reproduce, add logging, trace through service/ORM/jobs to root cause
• Backend performance optimization: identified SQL bottleneck and reduced API latency from seconds to milliseconds via indexing/join reduction/caching
• Proactive stakeholder/team alignment in unstructured environments using check-ins and early prototypes
• End-to-end ownership from problem definition through deployment and post-fix monitoring
• Deploying and integrating edge/embedded ML systems with multiple sensors on non-standard hardware (Synaptics SL1680)
• Systematic cross-layer debugging (hardware signals/protocols → Python services) using logic analyzer, kernel tracing, and timestamp logging
• Incremental integration/testing approach to isolate sensor/protocol conflicts in multithreaded environments
• Adapting ML deployment to hardware constraints (TFLite/YOLOv8 to .synapse conversion; Docker-based deployment)
• Improving robustness of noisy sensor pipelines via Python refactors (rolling median filter, confidence scoring, IMU consistency checks)
• Effective on-site collaboration with operators to reproduce real-world issues and deliver targeted fixes
• Built and deployed a production LLM automation agent for multi-step workflows across Notion, GitHub, and internal APIs
• Reliability engineering for agents via post-condition checks after every tool call (re-querying services to verify state)
• Regression detection using golden test scenarios to catch silent failures
• Designed custom orchestration for predictable, observable multi-step LLM/tool interactions (async, parallel calls, timed retries)
• Strong observability practices: logging plan, inputs/outputs, and decisions at each step for debuggability
• Systematic agent design approach: staged architecture (retrieval, planning, tool execution) with strict schemas and fallbacks
• Evaluation mindset with measurable metrics (tool success rate, grounding quality, error categorization) plus human-in-the-loop reviews
• Effective collaboration with non-technical stakeholders to deliver trusted, consistent LLM summaries with deterministic rule components