👋 Hi, I'm Rishikesh - an AI & Robotics Engineer building scalable intelligence systems for vision, language, and decision-making.
Designing simulation-driven AI systems and delivering reproducible, production-ready research that bridges lab to real-world gap.
- Visual learning and recognition, multi-view/3D vision, and robotics simulation
- Research-to-production: reproducibility, CI, datasets at scale, and reliable evaluation
- Cloud ML (GCP), containers, and tooling that makes experiments repeatable
- Expanding expertise in Large Language Models (LLMs) and Vision-Language Models (VLMs) for robotics - enabling natural language-driven perception, planning, and simulation workflows
- Preparing for Google Cloud's Professional ML Engineer certification by building ML projects on GCP
Programming Languages:
Python, C++, Bash, MATLAB, SQL
AI / ML Frameworks & Scientific Computing:
PyTorch, TensorFlow, Keras, HuggingFace, OpenCV, NumPy, Pandas, Matplotlib, SciPy, Scikit-learn
Robotics & Simulation:
ROS, ROS2, Gazebo, MoveIt, Isaac Sim, Falcon(Duality AI), Unreal Engine
DevOps & Cloud:
Linux, Git, Docker, Google Cloud, AWS, Jenkins
|
🏆 Team Lead | 8% Faster | 40% Cost Reduction NIST ARIAC 2023 - Agile Robotics for Industrial Automation Challenge |
🏆 20% Response Accuracy | 20% Latency Reduction LLM and RAG-based chat application with AlloyDB and Vertex AI integration |
|
🏆 93% Accuracy | 7% Faster | 12% Quality Boost Innovative approach to 3D Indoor Mapping and Object Segmentation for robot navigation |
🏆 IoU: 0.920 | VPQ: +15% | ViT Integration Motion prediction for Autonomous Vehicles using PowerBEV framework and Multi-Camera setup |
|
🏆 2.6x Compression Rate | 5,000+ Minutes Processed Advanced video compression and future prediction using GPT and VQ-VAE |
🏆 41 Stars | DQN + TD3 | Autonomous Navigation Autonomous vehicle navigation using RL techniques with DQN and Twin Delayed DDPG |
- Simulation-first mindset - design, validate, and iterate in high-fidelity environments before hitting real hardware, saving time and cost
- Research-to-production flow - every project ships with reproducible environments, fixed seeds, and exact dataset/weights versions
- Engineering discipline - CI/CD pipelines with linting, type checks, unit/integration tests, and automated benchmarks
- Storytelling through results - architecture diagrams, quantitative comparisons, and demo videos to make work easy to understand
- Scalable experimentation - containerized ML workflows on cloud (GCP) with distributed training and evaluation
- Clean handoffs - well-documented code, versioned changelogs, and minimal setup friction so others can run it end-to-end
- Toolsmithing - build or adapt tools (simulation assets, dataset converters, visualization utilities) to speed up team velocity
- Bias for impact - prioritize work that reduces friction for downstream teams and pushes projects closer to deployment
graph TD
A[AI and Robotics] --> B[Computer Vision / 3D]
A --> C[AI/ML]
A --> D[Robotics & Sim]
A --> E[DevOps & Cloud]
B --> B1[OpenCV]
B --> B2[PyTorch]
B --> B3[Geometry/SLAM]
C --> C1[HuggingFace]
C --> C2[Pytorch/TensorFlow]
C --> C3[Evaluation/Tracking]
D --> D1[ROS/ROS2]
D --> D2[Gazebo/Isaac/Unreal]
D --> D3[MoveIt]
E --> E1[Docker]
E --> E2[GCP/AWS]
E --> E3[CI/CD]