Ultra builds fixed-station, bi-manual AI robots that automate the highest-volume manual tasks inside e-commerce and 3PL fulfillment warehouses: order packing, sorting, and kitting. The thesis is a bet on an elite, twice-exited Harvey Mudd founding team applying imitation-learning robotics to warehouse order packing, the most labor-intensive, hardest-to-automate step in fulfillment. Rather than chase humanoids, Ultra fixes a two-armed robot on a roughly 5-by-5-foot base reaching across a 10-by-10-foot work cell, trading mobility for a cheaper, deployable-now system trained on example videos instead of explicit programming Web.
StoryHouse committed $50K from Fund I into the pre-YC $860K round on a $20M cap, sourced through cold LinkedIn outreach to co-founder Oliver Ortlieb in December 2023 and cultivated through the team's pivot from a robotics data-collection idea into warehouse manipulation Internal. Ultra went through Y Combinator's Summer 2024 batch Web. Per the deal record, Ultra raised a roughly $7M Seed extension in November 2025 on a $50M post-money cap SAFE, led by Physical Intelligence ($5M) with NextView Ventures ($1.5M): about a 2.5x markup on SH's entry cap, with the position marked Green Internal.
Warehouse automation is a secular growth market driven by e-commerce order volume, chronic warehouse labor shortages, and rising throughput demands on 3PLs Web. Order packing remains one of the least-automated, most labor-heavy steps, so a system that picks and packs variable items with example-trained AI attacks the part of the workflow incumbents have struggled to fully address.
| Player | Positioning | Funding / Stage | Edge vs. them |
|---|---|---|---|
| Ultra | Fixed-station bi-manual AI robot (OP1 “Operator”) for order packing, sorting, kitting in 3PL/e-commerce warehouses | Seed ext. ~$7M @ $50M post | — |
| Dexterity | Physical-AI industrial robots for palletizing and logistics | Late-stage, heavily funded | Ultra targets the packing cell specifically with a lower-cost, fast-deploy fixed station |
| Ambi Robotics | AI-powered parcel sortation and pack for shipping/logistics | Growth-stage | Ultra spans packing + kitting, not sortation alone |
| Covariant | AI piece-picking foundation model (Covariant Brain) | Heavily funded; core team acqui-hired by Amazon (2024) | Ultra owns the full station stack rather than a picking-only brain |
| Figure / humanoids | General-purpose humanoid robots | Mega-funded, multi-billion valuations | Ultra bets a fixed cell is cheaper and deployable today vs. general humanoids |
Moat: A data flywheel from example-trained manipulation in live warehouses plus a vertically integrated, purpose-built packing station, reinforced by a partnership with Physical Intelligence's robotic foundation models Web.
Likely acquirers are logistics and material-handling strategics and hyperscalers building physical-AI capability; Amazon's 2024 acqui-hire of Covariant's core team signals strong strategic appetite for warehouse-manipulation talent Web. On SH's $50K entry at a $20M cap, the November 2025 extension at a $50M post already implies roughly a 2.5x paper markup, with meaningful upside if Ultra reaches production-scale deployments Internal.
Miles Bird's pre-investment CEO interview with Max Friefeld, capturing metrics, milestones, and risks before finalizing the SH check.
CEO interview conducted by Miles Bird, 7/5/24, prior to finalizing SH's investment; company onboarded to Standard Metrics after the call.
Achievements to date are mostly recreating cutting-edge research results: they trained a vision-language-action model (trained with spoken language) to get a robot to pick up a specific object, and built the training infrastructure and OS pipeline. Funded by YC.
12-month milestones: customer-side, they target 3PLs as early customers, have had conversations with tech-forward 3PLs, and want ~10 customers to sign LOIs for a scoped system, targeting $10M in ARR in signed LOIs (~500 robots), plus development partners and early revenue or an R&D partnership. Engineering: move from a single miniature arm to a full-reach bi-manual V1 with finger end-effectors (design this summer, build in the fall), demo in Q1 2025, then convert to revenue. Fundraising in the fall to reach a core of ~10 people.
Risks: hiring for ML (not the founders' core expertise) though they just made their first hire; research progress, aiming to reach V1 quickly using published research; and data, capturing more of it early via tele-operation (analogy to a Waymo car with a safety driver). Plus some macro risks.
Metrics: net revenue $0, ARR $0, margins NA, retention NA, gross and net burn ~$50K each for the recent quarter (goal to stay under $50K/mo, may fluctuate with hardware), ~$835K cash, headcount 3 FT / 0 PT. Next raise: close by October (demo day end of September), minimum $2M, ideal $4–6M; cumulative raised ~$960K; fundraising strategy starts early August. Claremont: 3 Claremont employees; Claremont-affiliated investors/advisors Josh Jones and Vai. Asks: intros to hardware suppliers, hiring, and VC intros; future PhD scientist hire.
Call with Max on the pre-YC round mechanics and SH's rights, as the small round came together heading into YC.
Going into YC, the team didn't feel they needed to raise; the round was a way to get folks they'd been speaking to involved. Their Pioneer contact is former co-founder Patrick, who upped his check to $150K; an AI-fund HMC investor put in $100K; and two fund LPs who reached out put in ~$60K combined. The purpose of the ~$500K round is to be a little more hardware-focused (buy development hardware) and make a hire; with or without additional money, the YC funding runways them to December.
In response to SH's email, they asked questions about the MRL and pro rata and said they'd get back on whether they want to do this. Summary: good to go, though SH may not get MRL / pro rata / a VC interview slot.
Exploration call with Oliver as the team moved off a robotics data-collection idea toward AI-driven object manipulation, weighing logistics, retail, and home markets.
Previous exploration was a data-collection play for robotics; after ~30 conversations with researchers and roboticists they decided it was too early, noting even well-funded players (e.g. Figure, which had raised $700M at a $7B valuation for a humanoid) weren't pouring resources into data collection.
Current exploration: they bought a robot system to replicate research results. Most cutting-edge robotics research is on object manipulation, and a new flavor of ML lets a robot imitate a task after ingesting ~20 sample videos, a new paradigm feeding raw webcam footage into a neural net that outputs robotic-arm control. They want manipulation while moving and language-based control.
Business thinking favored a vertically integrated model owning the customer. Logistics and manufacturing are the most popular and most crowded; they were also open to retail; and leaning most into home automation, specifically elder home care, consumer tidy-up robots (~$5K target), companionship, and home monitoring/security. Plan was to focus on software first using off-the-shelf hardware, and to explore a ~$5K V1 “tidybot.”
Everything (UX, hardware, software, GTM, business model) will be hard; they wanted capability quickly to get interesting data quickly and would know within a month whether to press ahead. YC MFN kicks in July 9. Next step: a thoughtful email reminding them of SH's position. They might also seek a 4th co-founder doing AI/ML robotics controls and vision.