Who gets hired and onto which teams
A company that quintuples headcount in a year while landing FactSet, Fitch Solutions, and Microsoft Azure AI Foundry isn't just hiring. It's building an organization that can deliver LLM-powered workflows to the institutions that move capital. Hebbia recruits engineers and product thinkers who can build and ship LLM-powered tools at scale, offering competitive compensation grounded in market bands and a hiring process designed to assess real-world problem-solving under ambiguity. The company selects for intellectual rigor, ownership mindset, and fluency in turning research into product.
As of June 2026, according to Tracxn, Hebbia employs about 184 people, up from a fraction of that twelve months earlier. The org chart shows six departments: Engineering, Technical Operations, AI Strategy, Sales, Strategy and Operations, and Account Development. Each maps to a different slice of the problem Hebbia solves: turning unstructured financial documents into structured, queryable intelligence for asset managers, law firms, and corporate finance teams.
Engineering is the smallest named department at six people, but the titles tell the story: two senior engineering managers, a principal engineer, and three software engineers. This isn't a feature factory. The principal-engineer role signals deep systems work: retrieval architecture, evaluation pipelines, the plumbing that makes RAG reliable at scale.
The April 2025 appointment of Aabhas Sharma as CTO, charged with scaling infrastructure for the financial industry, confirms the shift from prototype to platform.
Technical Operations carries 20 people, the largest technical group.
At Hebbia that function owns the data pipelines, model evaluation loops, and deployment tooling that let the product ingest new document types, new jurisdictions, and new partner data feeds (Preqin, PitchBook, Third Bridge) without re-architecting the core. It's the operational counterpart to the research-heavy engineering team.
AI Strategy holds seven people. The name is deliberate: these aren't prompt engineers. They sit between the model layer and the customer workflow, designing the "patented interfaces" the company's own careers page highlights — the chain-of-thought orchestration, the citation architecture, the verification layers that make a lawyer or analyst trust the output.
First-party board listings for AI Strategist roles at the Principal and Strategic Accounts levels, with bands of $200k–$300k, confirm the seniority and client-facing weight of this function.
Sales (12), Account Development (5), and Strategy & Operations (7) round out the go-to-market side. The partnerships announced in the last twelve months (FactSet, Fitch, Azure AI Foundry, Third Bridge, PitchBook, Seyfarth) each require technical integration work, commercial negotiation, and ongoing account management.
That's where the Strategic Partnerships Manager and Partnerships Manager (Content) roles from the board listings slot in: they own the data relationships that differentiate Hebbia's corpus from a generic web index.
The FlashDocs acquisition, closed mid-2025, added workflow automation to the stack: document generation, not just retrieval.
That expands the engineering surface area toward agentic use cases: drafting memo sections, populating data rooms, automating diligence checklists. The org chart doesn't yet show a separate product team; product thinking appears distributed across Engineering, AI Strategy, and Technical Operations, a structure that works while the product is still defining its boundaries.
What emerges is a hiring profile split along two axes: deep systems engineers who can make retrieval deterministic at financial-grade accuracy, and AI strategists who can translate a partner's workflow into a repeatable product pattern. The company doesn't hire generalist "AI engineers." It hires for the specific fracture points where LLM pipelines break in production (evaluation, citation, latency, permissioning) and for the people who can sit across from a managing director at a top-50 asset manager and map their diligence process to a Hebbia workspace.
What it pays
Hebbia's compensation sits at the top of the AI-native market. Our board shows 22 salaried roles with a typical band of $90k–$274k and a median of $200k. Third-party aggregates cluster around that mark: Levels.fyi reports a company-wide median of $189k as of August 2026; Recruitingfromscratch puts the midpoint across 23 public postings at $215k. Glassdoor, drawing on 37 employee submissions, clusters Software Engineer total compensation at $173k, notably lower than the Levels.fyi Software Engineer median of $210k and the high-water mark of $337k, Levels.fyi's data shows.
The spread reflects real differences in level, equity grant timing, and whether a posting includes on-target earnings for commission-eligible roles.
| Role (first-party posting) | Base salary range (USD/year) |
|---|---|
| AI Strategist, Principal (NYC) | 225,000 – 300,000 |
| Strategic Partnerships Manager, Software (NYC) | 230,000 – 300,000 |
| AI Strategist, Strategic Accounts (NYC) | 200,000 – 275,000 |
| Solutions Engineer (NYC) | 170,000 – 260,000 |
| Partnerships Manager, Content (NYC) | 170,000 – 260,000 |
The first-party data skews toward go-to-market and deployed-engineering titles (AI Strategists and Solutions Engineers) because those are the roles Hebbia has posted most aggressively on our board. Levels.fyi fills in the engineering core: a Software Engineer median of $210k (base $160k, stock $100k/yr, bonus $0) with a reported ceiling of $337k. Project Manager sits at $179k; Recruiter at $189k. Only 9% of public postings explicitly reference equity, but the Levels.fyi breakdown makes clear that stock is the lever that pushes total compensation from the high-$100s into the $300k+ tier.
Equity follows a standard four-year vest: one-quarter at the one-year cliff, then roughly 2% monthly for the remaining three years. That schedule matches the market norm for Series B–stage AI companies and means a $100k/yr stock component delivers $25k at the cliff and about $8.3k/month thereafter. Candidates should model the grant value at the 409A price on their start date, not the preferred price from the last fundraise; the delta can be three to five times.
Benefits round out the package without fanfare. Medical, dental, vision, and mental-health coverage are standard. Parental leave runs four months for birthing parents, three for non-birthing (above the tech median). A $15k lifetime fertility benefit appears on the Ashby hiring page for Client Partner roles and is confirmed by Built In's benefits summary. Daily catered lunch, a DoorDash dinner credit, commuter support, unlimited PTO, and a 401(k) with no stated match complete the picture. Professional development (conferences, lunch-and-learns, job training) is listed but unquantified.
Negotiation leverage exists. Levels.fyi's own negotiators claim $50k+ upside on Software Engineer offers when candidates bring competing data. The gap between the Recruitingfromscratch median ($215k) and the Levels.fyi Software Engineer high ($337k) suggests the bands have room, especially for candidates who can demonstrate shipped LLM product work, the exact profile Hebbia selects for.
Inside the interview loop
Hebbia's interview loop runs four rounds over roughly three to five weeks, based on 31 candidate reports compiled by Dataford and updated weekly. The structure mirrors the company's high-velocity product cadence: a recruiter screen, one or two technical assessments blending live coding with system design, and a final round that often puts candidates in front of engineers shipping the core agentic infrastructure. Glassdoor's 28 reviews paint a similar picture: candidates most commonly rate the difficulty "medium," though experiences range from clearly stepped to chaotic, with some reporting unprepared or late interviewers.
The technical bar centers on object-oriented design (Dataford flags it at 100% coverage across reports) and the ability to reason from first principles rather than recite patterns.
Interviewers probe the "why" behind every data-structure choice and architectural boundary. A typical coding challenge might ask you to model a multi-agent orchestration layer: deciding which sub-agent triggers for a given task, tracking inquiry state across agents, and designing retry logic when an agent fails. Another track stresses real-time UX constraints: streaming partial results over Server-Sent Events or WebSockets, handling cancellation mid-stream, and keeping loading states honest during long-running document analysis.
A third vein tests complex rule implementation without spaghetti code (think Ultimate Tic-Tac-Toe constraints) and efficiency in data retrieval under latency budgets.
Must-have skills are explicit: fluency in Python or TypeScript, strong grasp of data structures and algorithms, and experience designing backend systems that handle asynchronous data streams. Nice-to-haves include prior LLM integration, document parsing and extraction, and startup-tempo shipping. But the differentiator isn't the checklist — it's how candidates navigate ambiguity.
Hebbia's own guidance tells interviewers to prioritize people who demonstrate deep understanding of their own code and can pivot when constraints shift mid-problem. The problems are grounded in the product: agents retrieving SEC filings, optimizing throughput on long-context windows, keeping the chat interface responsive under load. Collaboration is constant; engineers refine API contracts, cut latency, and ship behind feature flags daily.
The process moves fast once a candidate clears the technical baseline. That speed reflects a team that grew revenue 15× in 18 months, as Bloomberg Tech reported, while quintupling headcount, Bloomberg Tech's interview found, growth that demands hiring velocity without lowering the bar.
Candidates who advance tend to show three things: clean, maintainable code written under time pressure; system-design answers that address scalability, latency, and reliability as first-class concerns even in an AI context; and a product mindset that asks what the user actually experiences when a 200-page filing streams back token by token. The loop selects for engineers who treat research-grade problems as engineering work (not science projects) and who can own a slice of the platform from orchestration layer to UI streaming protocol without hand-holding.
Where the work happens
Hebbia operates as an on-site workplace. The company is headquartered in New York and maintains three office locations, with its roots planted in SoHo. A recent LinkedIn post from the company's own page marked "Hebbia's first day in our new office space" and noted the team was "sticking with our roots, working IRL in SoHo" while "preparing for serious growth." The language is deliberate: "IRL" signals a cultural commitment to in-person collaboration, not just a real-estate decision.
That commitment aligns with how the company describes its moment. In a Bloomberg Tech interview, CEO George Sivulka said, "I don't think it could be more exciting time to be building an AI company right here in New York." The phrasing — "right here" — reinforces that geography still matters for a firm building tools that sit at the intersection of finance, law, and large-language-model deployment.
Proximity to the asset managers, investment banks, and Fortune 500 companies that make up Hebbia's customer base (over a third of the top 50 largest asset managers by AUM, per the same interview) creates a feedback loop: product teams sit close to the workflows they're automating, and customer-facing roles can move between internal stand-ups and client sites without timezone friction.
The research does not detail the internal layout of the offices: no floor plans, lab configurations, or amenities lists appear in the available sources.
But the signals are consistent: a company that grew revenue 15× in 18 months is investing in physical space that can absorb rapid scaling while preserving the "IRL" density that early-stage AI product teams often rely on for the kind of ambiguous, cross-functional problem-solving the hiring process selects for. The new office in SoHo was framed explicitly as preparation for that growth.
With two additional locations beyond the New York headquarters, the footprint suggests a hub-and-spoke model rather than a single campus. The research does not name the other two sites or describe their functions. What is clear is that Hebbia has chosen not to decouple labor from geography — even as its product thesis argues that AI will decouple labor from humans.
The tension is noticeable: a company building "agent employees" meant to execute complex knowledge work autonomously still bets on human engineers sharing a room to ship those agents.
For candidates, the implication is straightforward. The interview process tests for systems thinking and coding fluency under ambiguity; the work environment then asks those same engineers to operate in person, alongside product, strategy, and deployment teammates who are translating research into production for financial-services clients. The office is not a perk — it's the default operating mode.
Who thrives here
The signals from Hebbia's own records (Glassdoor scores, Built In workplace data, and the company's public commentary) converge on a profile that is less about pedigree and more about how a person operates when the problem set is messy, the stakes are high, and the tooling is still being invented.
Eighty-eight percent of employees would recommend the company to a friend. Culture and values sit at 4.8 out of 5; career opportunities match that score. Work-life balance comes in at 4.2, a figure that holds even as revenue grew 15× in 18 months.
First, intellectual rigor that survives contact with enterprise reality. Hebbia does not build chatbots or single-turn search wrappers. The product is an analyst-grade agent that breaks complex finance and legal workflows into steps, executes them across multimodal inputs, and returns cited, auditable answers to Fortune 500 users. The company's own framing — "very few of them are able to do very much complex at all" — draws a line between prototype builders and people who can ship deterministic, secure, citation-backed systems into regulated workflows.
Second, an ownership mindset that maps to the OKR-driven, promote-from-inside structure. Built In notes "end-to-end ownership and OKR-driven goals creating clear advancement paths," and the company promotes from within, runs quarterly engagement surveys, and has an established employee-award cadence. The engineering team utilizes pair programming as a default, and open-door norms with in-person all-hands keep context dense across the 40-person product-and-tech core.
Third, collaborative fluency. Rob Huckin, hired as RVP EMEA after "months of rigorous due diligence," cited "a high-performance culture" as a decisive factor. The careers page emphasizes transparency, citations, and security as product requirements for a client base that includes major law firms and those firms.
Fourth, comfort building for buyers who will audit your work. The flexible work schedule and open office floor plan are enablers for deep work, not perks.
Finally, a growth trajectory that matches the company's. The board shows eight engineering roles, six AI/ML roles, and four sales roles open as of the latest Built In snapshot. The median salary band on Zero G Talent's board sits at $200k with a $90k–$274k range, with a density of $200k+ principal and strategist postings.
If your reflex is to optimize for local elegance over global utility, or to wait for a fully specified req before writing code, Hebbia will feel like friction. If your reflex is to own the outcome, instrument the ambiguity, and ship the thing the client can actually use — you'll be promoted, recognized, and asked to hire the next version of yourself.
Working in frontier tech? Zero G Talent tracks the openings: see every open Hebbia role, browse frontier tech jobs, the companies hiring, and the people building the field.