Lucen Robotics

Lucen Robotics is an early-stage robotics company founded by Linyan Fu and Apheth D'Almeida, building an autonomous restocking robot for retail stores. The company's approach is simulation-first: robots are trained and validated inside a realistic simulated supermarket environment before any physical deployment, using NVIDIA Isaac Sim to model store layouts, shelving, and product handling.¹ The team placed 2nd at the Night Hack competition hosted by Founders Inc in March 2026, splitting the prize pool and receiving a campus pass at the firm.³

Product

Lucen Robotics targets the night-shift restocking problem in supermarkets: after stores close, staff must manually check shelves for out-of-stock items and replenish them from backroom inventory. It is physically repetitive and labour-intensive work that most large grocery chains still do by hand. Lucen's system aims to replace this with an autonomous robot that can detect out-of-stock positions, navigate to backroom storage, retrieve the correct product, and place it in the correct shelf position — all without human direction.¹

The technical stack combines several components:¹

• Perception — A YOLOv8 computer vision model trained on real supermarket data for out-of-stock detection, achieving 84.25% SKU classification accuracy across 207 distinct products. Gemini AI is integrated in the pipeline alongside depth sensing.
• Navigation — SLAM-based (Simultaneous Localisation and Mapping) navigation inside the simulated store, enabling the robot to build and maintain a map of its environment as it moves.
• Simulation environment — NVIDIA Isaac Sim models realistic supermarket environments for training and evaluation. The simulation-first methodology allows the team to generate diverse training scenarios, vary store layouts and stock conditions, and measure repeatable performance without deploying physical hardware prematurely.
• Task execution — Pick-and-place restocking tasks: picking products from backroom inventory, transporting to target aisles, and placing items in correct shelf positions. The system includes basic failure recovery.
• Operator interface — A web-based dashboard providing real-time monitoring, inventory alerts, and analytics for retail operators.

Alongside the simulation-trained approach, Adrianna Lakatos — the Founders Inc investor who announced their Night Hack win — described the company's direction as "egocentric videos → robot training," referring to a technique where robots learn manipulation from first-person video of human workers performing tasks.³ This approach, sometimes called learning from egocentric observations, uses footage captured from a worker's perspective to train robot policies without requiring expensive teleoperation data.

Hackathon history

AMD Robotics and AI Hackathon, Tokyo (January 31, 2026). Before formalising Lucen Robotics, both founders competed in a 48-hour Physical AI hackathon in Tokyo alongside a team that also included Anish Senathi and Kashyap Kompella.⁶ The task was to train a robot to pick up a charger and plug it into a charging port — a contact-rich manipulation problem that is deceptively difficult in practice. The team used the LeRobot framework, an SO-101 leader/follower arm system, and trained an ACT (Action Chunking with Transformers) policy on a multi-view teleoperation dataset captured from three cameras (gripper, top, and side views).⁶ The robot could resume tasks after interruption and retry autonomously on failure. The team won a "World Intelligence Award" at the event.⁶

Lablab.ai — Launch and Fund Your Own Startup, Edition 1 (February 15, 2026). Fu and D'Almeida submitted Lucen Robotics to the lablab.ai hackathon focused on launching and funding AI startups.² The submission described the retail restocking system in full technical detail and is the earliest public documentation of the Lucen Robotics name and concept.

Founders Inc Night Hack (March 20, 2026). The company placed 2nd at Founders Inc's internal Night Hack hackathon, splitting a prize pool with 1st-place winner Enclave (secure data sandboxing) and 3rd-place Acepath (tennis coaching tracker).³ Adrianna Lakatos announced the result and awarded all three teams campus passes at Founders Inc for the month.³

Market context

The retail robotics market has seen multiple attempts at shelf-auditing and restocking automation. Simbe Robotics' Tally robot audits inventory by scanning shelves but does not physically restock; Brain Corp and others have built floor-cleaning and data-collection robots that are now common in large US grocery stores. None of the widely deployed systems complete the full restocking loop autonomously — picking a product from the backroom and placing it on a shelf. That physical manipulation task is the central technical bet of Lucen Robotics, and the simulation-first training methodology is how the founders are attempting to make it tractable at low hardware cost before physical deployment.¹

References

1. Lucen Robotics — AI Hackathon submission, Launch and Fund Your Own Startup-Edition 1 | Lablab.ai(accessed Apr 22, 2026)
2. Lucen Robotics team page — Robot for Restocking at Retail Stores | Lablab.ai(accessed Apr 22, 2026)
3. Adrianna Lakatos on X — Night Hack winners, Lucen Robotics 2nd place | March 20, 2026(accessed Apr 22, 2026)
4. Linyan Fu | LinkedIn profile(accessed Apr 22, 2026)
5. Apheth D'Almeida | LinkedIn profile