Leaf Jiang knows what it takes to make a laser judge distance. For 12 years at MIT Lincoln Laboratory, he designed coherent and direct-detection lidar systems: intricate machines built to send light into the world, collect what returns and turn elapsed time into a map. He made versions rugged enough for military platforms. He worked on a 15-channel laser Doppler vibrometer for robotic landmine detection, packing optics, lasers, electronics and suspension into an enclosure designed to tolerate shock, heat, dust and water. This was not a passing acquaintance with lidar. It was a long apprenticeship in the pleasures and punishments of precision.
Then Jiang founded a company around leaving the laser out.
NODAR, established in 2018 in Massachusetts, uses cameras and geometry to produce 3D depth. Its name carries the joke: no radar, no lidar. The technology is stereo vision, the familiar act of comparing two views of the same scene. The business is everything required to make that familiar act dependable on a vehicle that vibrates, warms, cools, splashes through weather and cannot politely stop while its sensors realign.
“This is a conversion not of time-to-distance, but angles-to-distance.”Leaf Jiang, AI in Automotive podcast
Jiang explains the distinction neatly. Lidar measures the travel time of light. Stereo measures angles. Find the same point in two camera images, compare its position and triangulate its distance. The schoolbook diagram is simple. The engineering diagram is a thicket.
The expert's reversal
Jiang came to that thicket with an unusual résumé. At MIT he earned a BS, an MEng and a PhD in electrical engineering and computer science. His early research touched optics and mode-locked lasers. At Lincoln Laboratory, his projects ranged from laser vibrometry to optical systems for studying the silhouettes of geosynchronous satellites. NODAR credits him with five patents, 32 academic publications and seven engineering awards before the company began collecting awards of its own.
The reversal from lidar to cameras was not a rejection of physics. It was an argument about the entire product. A sensor for a laboratory can be costly and rare. A sensor intended for mainstream vehicles must survive a supply chain and a bill of materials. Cameras already roll off production lines by the millions. They provide color and fine angular resolution. What they have traditionally struggled to provide is long-range, stable, metric depth when their mounting geometry shifts.
NODAR's central trick is to treat that shift as a live problem. Its cameras can be mounted independently, farther apart than a conventional stereo pair. A wider baseline improves long-range depth, but it also makes tiny alignment errors more consequential. Hammerhead, the company's platform, recalibrates the cameras frame by frame and then performs stereo matching in real time. The resulting depth map is deterministic geometry, not a monocular neural network guessing scale from learned examples.
This distinction matters because autonomy is full of seductive approximations. A system may recognize a truck yet be uncertain how far away it is. It may infer a road surface but miss the small debris sitting on it. Jiang has repeatedly focused attention on the 10-centimeter object at 150 meters, a compact way of expressing the highway problem: at speed, a machine needs enough distance to perceive, decide and act.
Five quiet years, then the microphone
At IAA Mobility in 2023, Jiang described NODAR as reaching an inflection point after five years of concentrated development and testing. The phrasing reveals the operating rhythm: disappear into the difficult work, then emerge with measurements. The same year, NODAR won the Grand Prix ACF in Paris and was named Overall Auto Sensor Company of the Year by the AutoTech Breakthrough Awards program. By then the company had raised a $12 million Series A led by New Enterprise Associates, with participation from Rhapsody Venture Partners.
The path from optical bench to venture-backed company had an intermediate stop. After Lincoln Laboratory, Jiang founded and ran Leaf Alden Consulting Group from 2015 to 2017. NODAR followed. That sequence matters because scientific authority and company building require different muscles. A founder must translate a physical principle into a recruiting pitch, a product schedule, a patent strategy and a reason for cautious automotive organizations to run a trial.
One early NODAR patent captures the translation. The non-rigid stereo vision camera system allows cameras to sit in separate locations while software corrects for movement between them. In ordinary stereo hardware, the camera pair is often locked into one rigid bar. That bar protects calibration but constrains where the cameras can go and how wide their baseline can become. Independent mounting opens space in the windshield, roof, mirrors or headlights. It also hands the software a relentless alignment problem. NODAR chose the freedom and accepted the problem.
This is where Jiang's background looks less like a credential and more like product architecture. The Lincoln Laboratory projects were not merely optical experiments. They demanded mechanical design, thermal management, electronics, software and tolerance for hostile conditions. A depth sensor on a production machine is similarly indifferent to departmental boundaries. Physics may begin the design, but the pothole gets a vote.
Capital brought a network. NODAR recruited advisers with histories at Daimler, Continental, Ford and BMW. It built support for NVIDIA compute platforms. It carried the stereo thesis beyond passenger cars into trucks, agriculture, rail, aviation, maritime systems and industrial monitoring. The common requirement is not a particular vehicle shape. It is a machine that must understand distance under conditions unfriendly to delicate instruments.
Jiang's public writing becomes livelier when the work leaves the road. In 2026 he opened a technical newsletter installment with a mock warning to his investors: he was about to reveal how NODAR builds ultrawide-baseline camera rigs. The assignment came from a customer that wanted to reconstruct the surface of water from roughly 50 meters away. Then the customer added night operation. Jiang's written response, complete with a smiling emoji, was that customer requests are always reasonable.
The team took the problem to a pool on a sunny day. Water is a hostile visual subject: reflective, changing and often short on the stable texture stereo matching likes. Night removes still more information. Yet this is exactly where a product thesis earns its keep. A geometry engine that works only when the scene behaves is a demonstration. A system that discloses its limits and keeps improving in awkward conditions might become infrastructure.
“There is an elephant in the room.”Jiang on the cost and performance constraints of autonomous-vehicle sensing
A mission measured in reaction time
Jiang's criticism of lidar is blunt, occasionally theatrical and grounded in manufacturing. He has pointed to cost, resolution, range and production complexity as reasons that laser systems may struggle to carry the full burden of mass-market autonomy. His alternative is not camera maximalism. NODAR describes sensor fusion options, and its software can sit alongside lidar and radar. The sharper claim is that stereo cameras can provide primary, dense 3D perception at camera economics.
That claim keeps dragging the company toward calibration. NODAR's 2026 research on 5-megapixel stereo sensing argues that more pixels improve range and density only when calibration accuracy rises with them. A one-meter camera baseline, the company's technical writing notes, is dramatically more sensitive to vibration than a compact pair. Resolution without alignment becomes an expensive way to produce prettier uncertainty.
The recent work also shows Jiang moving between founder and practicing engineer. He co-authored papers on high-definition 5MP stereo sensing and on depth error in fisheye stereo systems. He published walkthroughs of NODAR Viewer, the software window engineers use to inspect raw images, depth maps, point clouds, confidence and object tracking. The abstractions eventually land on a screen where someone clicks a pixel and asks a wonderfully severe question: how far away is that, really?
There is a personal through-line here, although Jiang rarely makes himself the spectacle. His career keeps returning to distance: laser pulses, satellite silhouettes, vibration, parallax, a scrap of road seen far enough ahead. The tools change. The demand for a trustworthy measurement does not.
His stated horizon is expansive. In an IAA interview, Jiang imagined autonomous systems spreading from cars to trucks, trains, boats, farms, robotaxis and air taxis. He tied that future to fewer collisions, less congestion and broader access to transportation. It is a founder's vision, naturally. It is also an engineer's: the future is allowed to be grand, provided somebody handles the calibration.
Jiang lives in Concord, Massachusetts, where NODAR is based. The company biography notes a wife and five children. That domestic fact sits almost comically beside a professional life spent measuring tiny angular differences at enormous distances. It also supplies the right scale for his ambition. Safer motion is not an abstract category when the people one loves are out on the road.
A useful lesson in Jiang's story is not that cameras will win or lasers will lose. Technology markets rarely grant such tidy verdicts. It is that contrarian companies can begin with unusually intimate knowledge of the incumbent. Twelve years of building lidar gave Jiang permission to see its compromises without caricature. NODAR is his proposed alternative, assembled from common cameras, uncommon calibration and the belief that a machine should know where the world is before it tries to move through it.