Careers at Triomics: Teams, Pay and How to Get Hired

The Hybrid Profile Triomics Hires

Triomics, a $50.8 million-backed (Alion's data shows) oncology AI company, runs a 54-person team (InsideOrg found) across San Francisco, New York, and Bengaluru — roughly half the roles remote or hybrid, with a posted compensation band of $92,000 to $203,000 and a $150,000 median. The roles don't fit a standard AI startup's org chart. The company sits at the intersection of oncology, large-language-model infrastructure, and enterprise healthcare deployment — each pulling the talent profile in a different direction. The person who gets hired demonstrates they can operate across the boundaries.

Triomics organizes hiring into four explicit lanes: engineering, clinical, go-to-market, and customer success. Engineering splits further. The backend product engineer owns services, application features, and the cloud infrastructure, deployments, and CI/CD that keep them running in production. The Forward Deployed ML Engineer builds and deploys AI agent pipelines that extract structured oncology variables from unstructured patient documents for pharmaceutical companies and cancer hospitals — full-cycle ownership from customer problem to production model. The DevOps/SRE role designs scalable, secure infrastructure and streamlines deployments. The Senior SDET and Software Tester roles own quality engineering and rigorous manual validation for systems where accuracy directly affects clinical decisions. The Director of Engineering in Bengaluru leads a team the company treats as a co-equal site, not a cost center.

Clinical and customer-facing lanes are equally technical. The Oncology Clinical Navigator owns adoption, training, and enablement at customer accounts, partnering daily with physicians, nurses, and coordinators — an internal role that lives in the clinical workflow. The Account Manager carries no quota; it is the single most informed person on account health, working with coordinators, physicians, registrars, and abstractors. The Technical Support Engineer, based in Bengaluru on U.S. shift hours, is the first dedicated support hire and acts as the eyes and hands of the engineering team when they are offline, monitoring production systems and triaging incidents.

The go-to-market lane shows one opening: a Senior Talent Recruiter tasked with scaling across customer success, forward-deployed engineering, GTM (provider and pharma), AI research, and other core functions — a role the description frames as "not a coordinator role" but a strategic partner to hiring managers. A Technical Program Manager sits at the intersection of engineering, product, and operations, driving cross-functional programs and bringing structure to ambiguity.

The company's own material (careers page, about page, job descriptions) emphasizes not a credential checklist but a set of operating preferences. "Care about the work" is phrased as the single best predictor of effectiveness: curiosity about cancer care, deep respect for clinicians, patience for healthcare's real constraints. "Build things end to end" means owning the outcome, not a slice, whether you are an engineer, designer, clinical lead, or GTM hire. "Write clearly" is called out because most internal communication is written and most external work is documents, papers, and citations. "Bias toward integrity over speed" translates to cited outputs, calibrated confidence, and saying "no" when "no" is the right answer — even if it slows the ship date. "Calm, durable pace" is explicit: long-cycle problem, hiring for multi-year relationships, pushing hard when it matters, not manufacturing urgency for theater.

Compensation: What the Board Shows

Triomics posts salary bands on open roles — a practice that lets candidates self-filter before investing time in a process that moves from first contact to offer in roughly three to five weeks for technical candidates. The live job board shows a $92,000–$203,000 base band with a $150,000 median across seven salaried postings as of September 2026. That range reflects a team that pays for specialized domain depth (oncology workflows, clinical-trial operations, regulated cloud infrastructure) rather than generic LLM prompting.

The board's six most recent postings, all dated within the last 90 days, give a clearer picture than any aggregate. Five are U.S.-based (four in New York, one remote-eligible); one sits in Bengaluru. I've converted the Bengaluru role to USD at the prevailing rate (~₹83/$) for side-by-side comparison.

Two patterns stand out. First, the forward-deployed ML engineer — a hybrid role that sits on customer sites at cancer centers like Memorial Sloan Kettering and Yale Cancer Center — commands the tightest, highest band ($170,000–$190,000). That reflects the rare combination of production ML engineering, clinical-domain fluency, and comfort working inside hospital firewalls. Second, the Director of Engineering role in Bengaluru carries a wide band (~$120,000–$240,000) that overlaps U.S. senior IC bands; the spread likely accounts for variance in scope (team size, P&L ownership) and the local market premium for leaders who can bridge U.S. product cycles with India R&D velocity.

Older third-party aggregates (Alion, Levels.fyi) show wider ranges — some estimated bands stretch to $279,000 for Platform Engineer or dip to $26,000 for ML/MLOps — but those figures are labeled "estimated" and predate current board postings. The first-party board data is narrower and more recent; treat it as the working range.

Equity is not disclosed on the board. The company has raised $50.8M total ($22M Series B (TechCrunch reported) in May 2026, $15M Series A (TechCrunch's data shows) mid-2024, $13.8M seed Dec 2022), so option grants exist but their strike price and percentage are negotiated per offer. Candidates should ask for the current 409A valuation and the option pool refresh cadence during the offer call — the board won't tell you.

Remote eligibility varies by function. The Account Manager posting explicitly lists "Remote (US)" alongside New York; engineering and ML roles are listed as New York onsite. The Bengaluru role is onsite at the Engineering/R&D office. If you need full remote, the Account Manager and potentially the Senior Talent Recruiter are your only current paths — though the company's "56% remote" figure for the New York office suggests flexibility increases after onboarding.

Bottom line: if your market rate sits inside $130,000–$190,000 base and you match the domain profile (clinical NLP, regulated cloud, or forward-deployed ML), the posted bands are realistic. If you're anchoring to $250,000+ base, this isn't the shop — yet. The tenfold ARR growth and fresh Series B mean the next compensation review cycle could shift bands upward, but today's posted numbers are the only firm data you have.

Inside the Interview Process

Triomics runs two distinct interview tracks — one for technical roles (engineering, data science, ML) and one for research-analyst and clinical positions — but both share a defining trait: speed. The company's career guides describe the process as "streamlined" and "efficient," with the technical track following the timeline noted above and research analysts seeing offers in two to four weeks. The guide notes the process is notably quick, so ensure your availability is clear and your preparation is front-loaded before your first conversation.

The opening screen

Every candidate hits a 30-minute phone or video screen with a recruiter or hiring manager. For technical roles, this is the "Initial Screening" — a fit check on background, motivation, and logistics. Recruiters verify you can work the required hours and that your salary expectations align with the band. Research analysts get an "Initial Conversation" that leans heavier on past experience with cross-functional teams, schooling, and any research-adjacent work. Availability comes up fast.

Technical rounds

Data scientists and engineers face two to three technical sessions. Expect live coding (Python or SQL), a take-home or onsite case study modeling a real oncology-data problem, and a system-design or ML-modeling discussion. Interviewers "are not just looking for 'right' answers, but for how you think, how you handle ambiguity, and how you communicate your reasoning." The guide is explicit: "Rigor here is non-negotiable. You must be able to design tests that yield actionable insights." Candidates who jump to code without clarifying the business context fail. "Clarify before you code: For SQL or case study questions, always ask clarifying questions to ensure you understand the business context before you start writing."

Behavioral rounds

One to two behavioral interviews follow, often with senior team members. The STAR method (Situation, Task, Action, Result) is recommended by the company's own prep material. Questions probe collaboration, conflict resolution, and how you've handled ill-defined requests. "Prepare for Ambiguity: You may be asked how you would handle an ill-defined research request. Show that you know how to ask clarifying questions to scope the work effectively." Candidates who can't articulate trade-offs between speed, accuracy, and complexity stall here. "Be ready to discuss trade-offs: In every technical solution, there are such trade-offs. Always highlight these."

Final assessment

The last gate is a conversation with a founder or department head. For technical roles it's labeled "Final Assessment"; for research analysts it may be the only round beyond the initial chat. The focus shifts to product sense, long-term alignment, and cultural fit. "Great candidates focus on the 'why' and the 'so what.' They don't just solve the technical problem; they explain how their solution impacts the business and what the potential trade-offs are." The hierarchy is visible — "Respect the Hierarchy: Always acknowledge the roles of those you are speaking with and understand how your research fits into the bigger company picture." Concision is rewarded: "Be Concise: When answering questions, get to the point quickly. Triomics interviewers appreciate candidates who respect the value of time."

What a strong application looks like

A resume that maps cleanly to the posted stack (Python, SQL, PyTorch/TensorFlow, cloud infra for engineers; clinical-trial ops, EHR data, regulatory familiarity for research/clinical roles) gets a screen. A tailored cover letter referencing Triomics' oncology-specific mission (automating unstructured EHR extraction for cancer centers) signals genuine interest. GitHub or portfolio links showing end-to-end ML projects (data ingestion → model → deployment) carry weight. Candidates who've worked in regulated environments (HIPAA, FDA, clinical research) move faster.

What disqualifies

Lying on the resume. Inability to explain your reasoning out loud. Treating the case study as a pure coding exercise without business framing. Vague answers on availability or location flexibility. Triomics hires for domain adjacency; cold pivots without a credible bridge get filtered at the first screen.

Three Sites, Three Functions

Triomics operates across its three primary sites (the same trio noted above) with a similar share of posted roles listed as remote or hybrid. The company's own materials and board postings make clear that each location carries a distinct functional weight, and candidates should understand the difference before committing to an onsite loop.

San Francisco: Strategy and Product Leadership

The registered headquarters sits at 601 Montgomery Street, Suite 1100, in San Francisco's Financial District. Highperformr.ai's company profile describes this office as driving "strategy and innovation," and the address appears on Craft.co as the single listed headquarters location. This is where founding leadership (co-founder and CEO Sarim Khan, per the Yale Cancer Center announcement) and product direction are anchored. The Montgomery Street suite is a commercial office building; there is no research lab or clinical space here. Candidates interviewing for product, strategy, or executive-track roles should expect their onsite loop to happen in this office, and the conversation will center on roadmap, partner integration (Epic, OncoEMR, iKnowMed), and the go-to-market motion across NCI-designated centers.

New York: Clinical Deployment and Customer-Facing Engineering

Triomics lists two Manhattan addresses on its contact page: 599 Broadway, 9th Floor (NoHo) and 1 World Trade Center, Suite 46C. LinkedIn's company page also shows "Headquarters: New York, NY." The Zero G Talent board reflects this concentration — five of seven recent salaried postings are New York–based: Site Reliability Engineer, Forward Deployed ML Engineer, Technical Support Engineer, Account Manager, and Senior Talent Recruiter. The Forward Deployed ML Engineer role in particular signals that New York is where model output meets clinical workflow: these engineers sit with oncology teams at Memorial Sloan Kettering, Yale Cancer Center, Mount Sinai, and the other eight-plus NCI-designated customers to integrate Prism inside Epic and OncoEMR. The Account Manager and Technical Support Engineer roles confirm a post-sales, implementation-heavy presence. Candidates for deployment, reliability, or customer-success tracks should prepare for an onsite at one of these two locations and expect interviewers to probe Epic App Marketplace approval experience, HL7/FHIR/CCDA fluency, and the ability to translate clinician feedback into product tickets.

Bengaluru: Core Engineering and Model Development

The India office is located at 145 HSR 6th Sector, Bengaluru South, Karnataka. Highperformr.ai states this office "supports software development and engineering," and the board's sole Bengaluru posting (Director of Engineering at ₹10M–₹20M/year) confirms senior technical leadership sits here. This is where OncoLLM, the longitudinal chart-reading model cited in Nature Digital Medicine (95% accuracy on trial matching) (PRNewswire reported), is built and iterated. The team works on the full stack: ingestion pipelines for PDF, TIFF, JPEG, and scanned faxes; citation and audit-trail infrastructure; and the agent framework that produces registry-ready abstractions across NAACCR, SEER, COC, and QOPI measures. Time-zone overlap with U.S. clinical partners is limited, so asynchronous communication and rigorous documentation are non-negotiable. Candidates for backend, ML, or data-engineering roles should expect a Bengaluru onsite (or a structured virtual equivalent) with deep technical screens on LLM evaluation, retrieval-augmented generation over longitudinal records, and production-grade MLOps.

Remote and Hybrid Reality

The board data shows the Account Manager role listed as "New York, NY, US / Remote (US)," and the aggregate (matching the earlier split) aligns with the spread: one Bengaluru role, five New York roles (one explicitly hybrid), and no San Francisco roles currently posted.

What the Geography Enables

The geographic split mirrors the product architecture: San Francisco sets the clinical-product contract (what oncology teams need before, during, and after the visit); New York delivers it inside the EMR and measures lift — 40% more trial matches, 30% enrollment lift, 67% faster chart review per ASCO presentations; Bengaluru builds the reasoning engine that reads hundreds of pages across notes, pathology, molecular, and outside records and cites every output line back to source. A candidate who understands which site owns which slice of that loop can tailor their interview narrative — and decide whether the commute, time zone, or travel expectation fits their life.

The People Who Last

The people who stay and ship at Triomics share a tolerance for clinical gravity. The company's own language ("ONCOLOGY IS HIGH-STAKES ITS AI SHOULD BE TOO") isn't marketing copy; it's a filter. Every model ships with citations baked in, every output traces back to a line in a patient's chart, and the team measures success in hours returned to coordinators and patients matched to trials who would have been missed. That standard selects for engineers and researchers who treat hallucination as a safety failure, not a metric to optimize around later.

The technical bar is shaped by the data itself. OncoLLM ingests hundreds of pages per patient (scanned faxes, pathology reports, molecular panels, years of longitudinal notes) and reasons across the whole record rather than chunking it page by page. That demands engineers comfortable with messy, multimodal ingestion (HL7, FHIR, CCDA, PDF, TIFF, JPEG) and researchers who can evaluate on clinical endpoints like 95% matching accuracy (published in Nature Digital Medicine) and 40% more trial matches (ASCO 2025), not just perplexity or F1 on clean benchmarks. The constellation-of-models architecture (not a single monolithic LLM) reflects a team that builds for cost-effective extraction at hospital scale, a constraint that came from the founders' own backgrounds: Sarim Khan's biotech research at MIT and clinical collaboration in Boston, and Hrituraj Singh's generative AI work at Adobe on multimodal understanding and controllable generation.

Customer proximity is non-optional. Triomics deploys inside Epic, OncoEMR, and iKnowMed, approved on the Epic App Marketplace, and works "side-by-side with our partner providers" at institutions including Memorial Sloan Kettering, Mount Sinai, and Yale New Haven Health. The company's stated principle ("Healthcare AI gets built with healthcare") means forward-deployed engineers and support staff sit in tumor boards, watch coordinators screen patients, and iterate on workflows that clinicians actually use. Sebastian Rhodes, who leads strategy and operations, said the mission is aligning generative AI advancements with oncology's daily reality. Candidates who need a clean API boundary between product and user will chafe; the ones who thrive treat clinical feedback as the primary signal.

The culture signals are mixed and instructive. Glassdoor reviewers call out "intelligent, adaptable, and compassionate colleagues" and "supportive team culture," and AmbitionBox rates work culture 4.4 out of 5 across 11 reviews. But the same Glassdoor source notes "concerns about a lack of cohesive culture and communication issues" — a pattern common in early-stage teams scaling across its three sites with that same remote-or-hybrid split. The compensation band ($92,000–$203,000, median $150,000) and the investor roster (Battery, Lightspeed, Nexus, Y Combinator) signal a well-capitalized, high-expectation environment. The people who last are the ones who can operate in that ambiguity: they ship auditable, cited outputs, incorporate a coordinator's workflow feedback, and don't wait for a process doc to tell them the next step.

Working in AI? Zero G Talent tracks the openings: see every open Triomics role, browse AI jobs, the companies hiring, and the people building the field.