Hark

Hark is an AI company building personalized, multimodal intelligence across speech, text, vision, persistent memory, hardware, and agentic computer use.

Company Signal

Hark presents itself as a company building advanced personal intelligence: systems that can listen, speak, see, remember, act proactively, and interact with the world through new hardware and agentic computers. Its public site frames the product as a new interface to AI, not a chat-only app. Source: https://www.hark.com/, 2026-06-30

Business Wire's March 2026 launch release says Hark is building foundation models, software systems, native hardware, and new interfaces together from the start. The target capabilities are multimodal intelligence, personalized memory, proactive behavior, real-time speech, and native hardware that can listen, see, and interact with the world in real time. Source: Business Wire, 2026-03-24

In May 2026, Hark announced a $700M+ Series A at a $6B post-money valuation, led by Parkway Venture Capital with participation from NVIDIA, AMD Ventures, ARK Invest, Intel Capital, Qualcomm Ventures, Salesforce Ventures, Brookfield, Greycroft, Prime Movers Lab, Align Ventures, and Tamarack. The capital is positioned for GPU infrastructure, AI model development, team growth, and AI-native hardware. Source: Intel Capital, 2026-05-21; Hark X post, 2026-05-21; Brett Adcock X post, 2026-05-21

The public careers board reinforces that positioning. Open departments include Product Engineering, Computer Use Agents, AI Foundation Models, AI Infrastructure, Embedded Software, Hardware Engineering, Privacy and Security, and Design. Product Engineering roles are concentrated in San Jose and include Backend Engineer, Frontend Engineer, Full-Stack Engineer, Platform Engineer, Integration Engineer, and mobile roles. Source: https://job-boards.greenhouse.io/hark, 2026-06-30

The careers board listed 48 open roles on 2026-06-30. Department counts were: Hardware Engineering 10, Product Engineering 8, Embedded Software 7, AI Foundation Models 6, Design 3, Privacy and Security 3, AI Infrastructure 2, Computer Use Agents 2, plus AI Safety, AI Data Operations, Program Management, Finance, Supply Chain, Marketing, and People. Source: Greenhouse board API, 2026-06-30

Team And Design Thesis

Business Wire reported more than 45 researchers, engineers, and designers at launch. TechCrunch later reported around 70 employees in May 2026 and said funding would support hiring in hardware, product design, and AI research. Source: Business Wire, 2026-03-24; TechCrunch, 2026-05-21

Abidur Chowdhury, formerly an Apple industrial designer associated with iPhone Air, leads design. TechCrunch's useful design signal is that Chowdhury framed future UX around finding the right thing for each individual, while avoiding interfaces that put an awkward layer between people and the world. Source: TechCrunch, 2026-03-24

The product thesis appears to be "ordinary-person agent harness": an assistant powerful enough to plan and act, but presented through a hardware/software interface that does not require users to understand prompt engineering, tools, or agent internals.

Computer-Use Agent Signal

Brett Adcock publicly described Hark's computer-use-agent goal as "use any computer as well as a human" and called the target a "digital humanoid" that can navigate the internet. Tanmay Gupta joined Hark to shape the CUA effort and described it as personalized intelligence that brings frontier-level task planning and execution to everyday goals. Source: Brett Adcock X, 2026-06-26; Tanmay Gupta X, 2026-06-26

That makes computer use a central Hark signal, not a side feature. For interview prep, the strongest technical framing is: browser/control agents force product engineering to solve state recovery, idempotency, permissions, confirmation, observability, and DOM fragility.

Product Engineering Signal

The Product Engineering roles make the interview signal fairly explicit:

• Frontend work centers on real-time AI UI: streaming interactions, partial responses, tool events, out-of-order state updates, accessibility, performance, and polish. Source: Hark Frontend Engineer role, https://job-boards.greenhouse.io/hark/jobs/4193733009, 2026-06-30
• Full-stack work centers on product surfaces across voice, text, and vision, with agent workflows that combine tool use, memory retrieval, and multi-step reasoning. Source: Hark Full-Stack Engineer role, https://job-boards.greenhouse.io/hark/jobs/4185889009, 2026-06-30
• Backend work centers on high-concurrency services, low-latency streaming, state management for long-running agent workflows, tool execution, memory retrieval, sandboxed execution, observability, and evaluation. Source: Hark Backend Engineer role, https://job-boards.greenhouse.io/hark/jobs/4193725009, 2026-06-30

The shared product thesis is that model capability only matters when it becomes fast, coherent, trustworthy user experience. That maps directly onto Code Taste and Security and Review Skills: explicit state, clear contracts, deterministic UI behavior around nondeterministic models, and visible proof that the thing works.

Kevin Fit

Hark is a strong narrative fit for Kevin because it intersects three existing lanes:

• Dedalus Labs and Agent Machines: production agent infrastructure, MCP, tool execution, sandboxes, and agent runtime thinking.
• Kevin-Wiki: persistent memory, agent-authored knowledge, skills, automations, and "agents should compound" as a lived operating system.
• Product/frontend taste: Hark's product roles repeatedly emphasize latency, trust, multimodal UX, and polish, which are already standing parts of Kevin's UI/product defaults.

For recruiting calls, the concise positioning is:

"I have been building production agent infrastructure professionally and using agents as my own development environment. Hark is interesting because the hard problem is turning model capability into an interface people trust."

Interview Implication

For an app-building coding interview where AI and the normal environment are allowed, Hark is likely evaluating the full product loop: understand the ask, choose a narrow app shape, use AI without surrendering judgment, ship the vertical slice, keep state coherent, and talk through tradeoffs.

See Hark App-Building Interview Prep for the live round plan. See Hark Live Sidepanel for a call-side brief with high-signal talking points.

---

Timeline

• 2026-07-01 | Project graph refresh checked this page against Kevin's live GitHub inventory and PortfolioMon scrape. No direct repo/card match was promoted here, so the page remains a context or umbrella project page. Source: raw/github/kevin-repos-2026-07-01.json; raw/sites/kevinliu-biz-2026-07-01/portfolio-projects.json

• 2026-06-30 | Created from Kevin's pasted recruiter note that the Hark coding interview allows the normal environment and AI, involves creating an app, and requires laptop/environment readiness. Public Hark sources were checked for product context: personal intelligence, multimodal interfaces, Product Engineering roles, streaming AI UI, memory, tool use, and production agent runtime concerns. Source: User, 2026-06-30; https://www.hark.com/; https://job-boards.greenhouse.io/hark

• 2026-06-30 | Expanded with deeper public-source research: launch and funding details, team/design thesis, Abidur Chowdhury, 48-role Greenhouse board scan, repeated careers signals, Brett Adcock's computer-use-agent framing, Tanmay Gupta's CUA role, and the resulting interview/product-engineering implications. Source: Business Wire, Intel Capital, TechCrunch, Greenhouse, X, 2026-06-30