Embodied AI Engineer

**Company Overview** The world is moving towards an autonomous future where sooner or later everything that is driven, pushed or pulled today, will be moving by itself. At Flo Mobility, we believe that the true adaptation of autonomy lies in the seamless coexistence of humans and bots. By leveraging the power of sensors, computer vision, edge computing, and actuators, we optimize this relationship to deliver the best speed, agility, comfort and control. Our hero product, FLO Hauler, is an Autonomous Electric Material mover, helping customers reduce time and cost of moving material across Construction, Agriculture, Mining and Warehousing industries. Flo Hauler is being used by industry leaders like L&T, PSP, Total Environment to gain 50% savings on cost and time for the last mile of material movement. Our solutions are offered on RaaS (Robot-as-a-Service) model. The no capex approach makes it easy for our clients to adopt our MMRs (Material Movement Robots). **Required experience:** 0 - 4 years (freshers with research experience welcome) **Job type:** On-site, full time **Role Overview** We are seeking an Embodied AI Engineer to bring an AI-first perspective to our robotics team. Our robots work in unstructured, outdoor environments, and we want learned models, foundation models and AI agents to drive how they see, behave and improve. You will move fast from idea to working prototype: build AI agents, create large datasets with automatic labelling, train perception models and teach robots new behaviours from data. You will also keep the team up to date with the latest research and turn the best of it into product. **Responsibilities** 1. AI-First Thinking: Find where AI can replace or improve hand-built parts of the system, and prove it quickly with prototypes and numbers. 2. AI Agents: Build quick, useful agents with LLMs and VLMs (tool calling, retrieval,multi-step workflows) for the team, operations and the robot. 3. Datasets at Scale: Build pipelines that turn raw camera and LiDAR recordings into large, curated, searchable datasets. 4. Automatic Labelling: Label large datasets automatically with foundation models (open-vocabulary detection, segmentation, VLMs), with human review only where it is needed. 5. Perception Models: Train and fine-tune vision models for scene understanding on real-world robot data, and run them on edge hardware. 6. Behaviour Learning: Train robot behaviours with imitation learning and reinforcement learning, from demonstrations and in simulation. 7. Evaluation: Set up benchmarks and metrics so every model is measured, and a new model ships only when it is better. 8. Research to Product: Follow the latest research, reproduce promising papers quickly and share what you learn with the team. **Education Background** 1. B.E./B.Tech., M.E./M.Tech., M.S. or Ph.D. in Computer Science, AI / ML, Robotics, Electronics, or a related field (Mandatory). 2. Freshers with strong research experience (thesis, publications, research internships or open-source work) are encouraged to apply. 3. Mandatory: Must Have Skills And Hands-On Experience 4. Strong Python and PyTorch. 5. Hands-on training of deep learning models (vision, language or robot policies) through work, research or projects. 6. Hands-on with LLMs or VLMs: prompting, fine-tuning or building agents. **Skills Required For The Job Role** 1. Foundation Models: Vision-language models and open-vocabulary detection /segmentation (e.g. SAM, Grounding DINO, Qwen-VL). 2. AI Agents: Building LLM agents with tool / function calling, retrieval and agent frameworks (e.g. MCP). 3. Data Engineering: Dataset curation, auto-labelling, active learning and data versioning at scale. 4. Robot Learning: Imitation learning, reinforcement learning and vision-language-action models (e.g. OpenVLA, pi0, SmolVLA, LeRobot). 5. Simulation: Training and testing in simulators (e.g. Isaac Sim / Isaac Lab, MuJoCo, Gazebo). 6. Deployment: Model export, quantization and inference on edge GPUs (ONNX, TensorRT). 7. Software: Linux, Git and clean, tested code. **Good To Have** 1. Publications at robotics or ML venues (ICRA, IROS, CoRL, RSS, CVPR, NeurIPS) or strong arXiv work. 2. ROS2 experience. 3. 3D perception: point clouds, LiDAR-camera fusion. 4. Synthetic data generation or world models. 5. Distributed training on cloud GPUs. 6. Open-source contributions in AI or robotics.