The Calibration Bottleneck, Automated
Conductor Quantum, a four-person startup from Y Combinator's Summer 2024 batch, has built software that collapses qubit calibration from weeks to minutes. Quantum engineers still spend weeks coaxing a single qubit into coherence, adjusting voltages and magnetic fields one parameter at a time, then repeating the process when the device drifts overnight. That manual loop — measure, analyze, decide, repeat — has become the field's hidden bottleneck.
The company's flagship product, Control, sits at the hardware layer of a two-part stack. Above it, Coda provides a natural-language interface that lets researchers describe an experiment in plain English and receive executed results. Below, Control connects to the control electronics of any major qubit platform — superconducting, spin, trapped-ion, or photonic — and runs the full calibration autonomously. The workflow is explicit: define the device in a configuration file mapping instruments, channels, and qubit parameters; Control then drives the hardware through characterization experiments (spectroscopy sweeps, Rabi oscillations, resonator readout), feeds raw data into machine-learning models trained on quantum measurement data, and uses the model outputs to decide the next action: adjust a parameter and re-measure, compensate for crosstalk, or advance to the next calibration step. The loop repeats until every parameter is optimized.
Brandon Severin, Conductor's CEO, lived the problem during his PhD at Oxford's Natalia Ares Group: "It took me weeks to just get like halfway along that journey… I said to my supervisor… never again. This is crazy." His co-founder and CTO, Joel Pendleton, had reached the same conclusion working across carbon nanotubes, superconducting circuits, and silicon spin qubits. Severin and Pendleton met at Oxford during their PhDs on AI for quantum computing. They founded Conductor in 2024 with a shared conviction that the only path to useful quantum computers — machines with millions of qubits — required automation from day one. The company's own figures put the acceleration at up to 1,000×: qubits created and tuned in minutes rather than weeks, letting teams iterate and experiment in hours.
Control's architecture is built for that scale. A parallelized design with no central bottleneck allows the software to calibrate tens of qubits today and, in principle, billions tomorrow. ML models classify qubit states and detect anomalies in milliseconds, replacing hours of manual inspection. The system is hardware-agnostic by design: it integrates with the control electronics a lab already owns. Early traction came from a collaboration with Finland's SemiQon, combining SemiQon's silicon hardware with Conductor's AI control stack to demonstrate autonomous operation on semiconductor chips.
In a traditional lab, calibration cadence is limited by human attention and sleep schedules. At 1,000 qubits, manual recalibration becomes mathematically impossible — the hours do not exist. Conductor's approach shifts the constraint from people to compute. The company's recent demonstration with EeroQ and NVIDIA's Ising models showed an AI agent running an electron-trapping protocol on real hardware from a single plain-English prompt, logging every parameter sweep and measurement without a researcher at the controls. That proof of concept, published May 2026, points to autonomous labs where software handles the overhead so researchers can do science.
Hiring for the Hybrid Engineer
Conductor's calibration breakthrough rewrote its hiring plan. The founding team that closed a $500K seed in May 2024 now advertises five distinct "Member of Technical Staff" roles on Y Combinator's job board: Quantum Engineer (Ion Trap), Quantum Engineer (Neutral Atom), AI Systems Engineer (Quantum), Quantum Engineer (Superconducting), and Quantum Engineer (Spin). Each listing carries the same San Francisco base and the same compensation band: $120,000 to $180,000 annually plus 0.25% to 1.50% equity. A LinkedIn posting for the Neutral Atom role, live as of April 2026, showed 89 applicants after four months: evidence that the hybrid profile they're chasing exists, but isn't abundant.
| Role | Base Salary | Equity |
|---|---|---|
| All five MTS positions | $120K–$180K | 0.25%–1.50% |
The job descriptions make the hybrid requirement explicit. A Quantum Engineer at Conductor doesn't just tune qubits; they "shape Conductor's core offerings: AI software that controls quantum hardware" and "build and maintain the first natural language interface for quantum computing, suitable for both human researchers and AI agents." The AI Systems Engineer role mirrors that scope from the model side: agentic quantum algorithm discovery, circuit generation tooling, and the calibration feedback loop that ties model output to pulse-level hardware behavior. The company's recruiting pitch frames it bluntly: "Today, AI generates production code faster and more reliably than most of the world's software engineers. Code is math; math is an instruction set; an instruction set is a quantum circuit. AI is already generating quantum circuits. It just needs reliable hardware to run them on."
That thesis sets a high bar. The Neutral Atom listing asks for "at least one calibration improvement in production that has moved a key metric," citing state preparation and measurement fidelity, time to a defect-free array, and homogenized trap depths as examples. It also expects the hire to "lead neutral atom operations at Conductor" and "drive the control and calibration roadmap, including which systems we bring onto the platform next." In other words, they want engineers who have already closed the loop between software automation and physical qubit performance, not theorists who have only simulated it.
Conductor's Y Combinator job page lists "Similar Jobs" from Torus, Aseon Labs, Steinmetz, Nine Fives, Simantic, Adialante, Charge Robotics, Paces, and Proception Inc (a cluster of early-stage startups all fishing in the same pond). Larger players are circling too: NVIDIA's Ising models and Microsoft's Majorana 1 roadmap imply demand for engineers who can translate AI-generated circuits into calibrated hardware runs. Conductor's edge is specificity (its software already targets spin and superconducting systems with a chat interface), but that edge only holds if the team can keep pace with the hardware roadmap.
The hiring signal extends beyond Conductor's headcount. The National Quantum Initiative's 2024 reauthorization expanded workforce funding. Universities are adjusting curricula, but the lag between degree programs and production-ready engineers is measured in years. The hiring surge is the direct consequence of collapsing a weeks-long manual process to minutes: the new bottleneck is the people who can keep that automation honest at scale.
Competitors React: The Control Layer Commoditizes
Conductor's demonstration of autonomous calibration has landed in an industry already racing toward automation, and the reaction reveals how quickly the baseline is shifting. IonQ, the trapped-ion leader now public on NYSE, has published on automated circuit optimization since 2018 and optimal gate calibration since 2021, work that fed directly into its 99.99% two-qubit gate fidelity claim and a roadmap targeting 2 million physical qubits by 2030. But IonQ's approach has historically been vertically integrated: its own hardware, its own control stack, its own error mitigation. Conductor's pitch (hardware-agnostic software that drops into any silicon-spin or superconducting lab) threatens to commoditize the calibration layer that IonQ and others have treated as a moat.
A clear signal came in May 2026 when Conductor and EeroQ demonstrated an autonomous quantum lab workflow on NVIDIA's Ising AI models (a demo Conductor published showing an AI agent running an electron-trapping protocol from a plain-English prompt). NVIDIA's Ising launch, framed as "the world's first open AI models to accelerate the path to useful quantum computers," positions GPU-accelerated ML as the universal control plane. For Conductor, the partnership validates its parallelized, bottleneck-free architecture; for NVIDIA, it proves Ising can drive real hardware, not just simulation. Conductor also released CODA MCP: a Model Context Protocol server that lets AI agents call quantum computing resources as native tools. That move reframes calibration from a specialist's craft into an API call, and it forces every full-stack quantum vendor to decide whether to build, buy, or expose their own control layer.
Funding flows tell the same story. Pasqal, the neutral-atom competitor, closed at least €340 million (roughly $370 million) in late 2024 ahead of a planned dual Nasdaq-Euronext listing (a war chest explicitly aimed at challenging IonQ and Rigetti on scale and software integration). Microsoft itself unveiled Majorana 1, a topological-qubit processor that, if it scales, would rewrite the error-correction calculus entirely. None of these moves are direct responses to Conductor's four-person team, but they share a logic: the bottleneck has shifted from qubit fabrication to qubit control, and whoever automates control first sets the pace for everyone else.
The UK National Quantum Computing Centre's 2024 hackathon offered a proxy for developer sentiment. For the second consecutive year, every winning team ran on IonQ hardware using Classiq's software stack: evidence that the IonQ-Classiq abstraction layer still wins mindshare. But Classiq CEO Nir Minerbi noted the integration with IonQ "enabled them to achieve these winning results," a phrasing that hints at growing demand for plug-and-play control. If Conductor's CODA MCP gains traction, the next hackathon could see teams switching backends as easily as they swap Python libraries.
For hardware vendors, the strategic choice is hardening: double down on proprietary control (IonQ), open the stack to third-party automation (potentially Pasqal), or build an AI-native control layer in-house. Conductor's seed funding is modest against nine-figure rounds, but its leverage is architectural: software that scales from tens to billions of qubits without rewriting the calibration loop. The industry's response suggests architecture matters more than headcount.
Scaling Toward a Million Qubits
The calibration bottleneck that Conductor automates is only the first of several that must fall before quantum computers reach the hundred-thousand-qubit scale where error-corrected algorithms become practical. Caltech physicists demonstrated a 6,100-atom neutral-atom array in September 2025 (a leap from the previous hundreds) while maintaining 99.98 percent single-qubit fidelity and 13-second coherence. The same research notes that robust quantum computers will require hundreds of thousands of qubits. Neutral-atom experiments have grown at roughly 1.8 times per year over the past decade with gate errors falling by a factor of 0.6 annually, though the roadmap coalition cautions these are trends, not guarantees.
Conductor's architecture is explicitly built for that trajectory. The company's documentation states its control software is "designed from the ground up to operate large-scale quantum computers, from tens of qubits to billions" with a "parallelized architecture with no bottlenecks." That claim will be tested as platforms push toward the 100,000-qubit regime where the neutral-atom roadmap identifies laser power and integrated photonics as the next hard constraints. Current arrays above 3,000 rubidium atoms already demand approximately 15 watts at 850 nanometers; kilowatt-class commercial lasers could support 100,000 atoms but introduce heat, coating damage, and noise risks. PASQAL, a coalition contributor, targets gate rates above 100 cycles per second by 2028 through silicon nitride photonic integration, up from roughly one cycle per second today. Conductor's software will need to co-evolve with that photonic layer, managing not just calibration but real-time crosstalk mitigation and readout optimization across thousands of parallel channels.
| Milestone | Current | 2028 Target |
|---|---|---|
| Neutral-atom gate rate | ~1 cycle/sec | >100 cycles/sec |
| Array size (rubidium) | ~3,000 atoms | 100,000 atoms |
| Laser power | ~15 W | Kilowatt-class |
Engineers must understand pulse-level quantum control, machine-learning-based system identification, and the compiler stack that schedules both logic gates and (for movable-atom architectures) physical qubit transport. The neutral-atom roadmap highlights a compiler problem with "no good classical analogue": scheduling atom movement, travel speed, parallel operations, and real-time response to atom loss and error signals. Conductor's job postings reflect this hybrid profile, seeking "cracked engineers whose sole focus is to develop software to enable fault-tolerant quantum computation." The company's CODA MCP server, which lets AI agents call quantum resources as native tools, signals a second skill vector: building the abstraction layers that let algorithm developers work without hardware expertise.
Funding trajectories align with the technical milestones. The broader field benefits from the reauthorized U.S. National Quantum Initiative, which expanded funding in 2024 and includes dedicated workforce-development provisions. Private capital follows similar markers: Pasqal's €340 million raise, IonQ's 2-million-qubit target, and QuEra's demonstration of 96 error-corrected logical qubits on 448 physical atoms in January 2026. Each milestone raises the bar for the next funding round, and the startups that survive will be those whose control stacks scale without linear growth in human operators.
The neutral-atom roadmap projects quantum utility within 10 years if the field reaches 100,000 to one million physical qubits with high-fidelity control. That projection sets the hiring window. Conductor's four-person team as of 2024 must grow into the discipline that writes the operating system for that machine. The engineers who join now will not just tune qubits; they will define the APIs, the error-correction interfaces, and the compiler heuristics that determine whether silicon spin qubits or neutral atoms or superconducting circuits win the scale race. The calibration automation that sparked this hiring wave is the prerequisite. The real product is the control plane that makes a million-qubit machine programmable.
The loop that once took weeks — measure, analyze, decide, repeat — now runs in minutes, unattended. The engineers Conductor is hiring today will write the software that keeps that loop honest at a million qubits.
Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.