Fin Interview Prep

Company facts, interview logistics, and what to expect for the Fin Staff/Senior Machine Learning Scientist onsite.

Fin (formerly Intercom) is being acquired by Salesforce while operating independently

Founded 2011 as Intercom; rebranded to Fin in May 2026 (company and flagship AI product now share the name)
Salesforce agreed to acquire Fin for approximately $3.6B in June 2026; expected to close Q4 FY2027
Fin operates independently until the deal closes
CEO: Eoghan McCabe (co-founder). President: Archana Agrawal
Scale: 30,000+ companies, crossed $100M ARR in 2026, growing 3.5x
Flagship metric: 67% average resolution rate (verified positive resolutions), top performers 80-84%

Model: Fin Apex 1.0, post-trained on proprietary production support data, replacing foundation models
Vs Claude Sonnet 4.6 in production: +2.8% resolution rate, 0.6s faster latency, 65% fewer hallucinations
Pipeline: 7 discrete models: language detection, issue summarization, knowledge retrieval, result reranking, response generation, feedback parsing, escalation routing
Retrieval: RAG based, retrieves ~40 candidate documents, reranks, filters context before generating
Hallucination mitigation: grounded in retrieved context (not pretraining knowledge), a dedicated hallucination checker at generation time, a final AI Engine safety/accuracy check
Published research: actor-critic approaches to reduce hallucinations, structured agentic RAG for ecommerce, topic modeling for dialogue

Source: Fin's official candidate interview schedule. Total time: 2.5 hours.

Confirmed (Fin's schedule doc): evaluates communication effectiveness, knowledge of fundamental data structures, and general problem-solving approach
Inference, not confirmed: given Fin's domain, problems may lean toward retrieval, ranking, or routing/escalation logic (top-k selection, priority queues, graph traversal)
Prep focus: narrate your reasoning, state time/space complexity, cover edge cases, then connect to a domain angle if one is offered

Confirmed (Fin's schedule doc): rapid-fire questions across different aspects of ML
Prep focus: breadth over depth. Crisp, correct answers on fundamentals (bias/variance, overfitting, regularization, evaluation metrics, common architectures)
Practice giving a complete answer inside 3 minutes per topic

Confirmed (Fin's schedule doc): candidate presents a slide deck about a project felt to be impactful at a previous role, showcasing ML skills
Must bring the deck and be ready to share screen in advance
That deck is a separate project: see the fin-project-review deck for the actual content

Senior: project-level ownership, deep single-domain competence, strong execution and fundamentals
Staff: cross-team or org-level scope, sets technical direction, navigates ambiguity and tradeoffs
Staff: decisions stay defensible under hindsight, impact is quantified rather than only described technically
Staff: communication and organizational influence carry as much weight as technical depth

Background for shaping the fin-project-review deck. Not a talking point to raise directly in the interview.

Values Assessment Session (45 min): impact at Intercom/Fin, how you would live company values
ML Skills (60 min): deeper ML-specific knowledge and skills across a series of topics
Technical Design (60 min): role-play translating product requirements into an ML design
Lower priority for now, later stage; noted here for awareness only