Full-Time
AI Infrastructure Engineer
Updated on 8/20/2026
Fin
1,001-5,000 employees
AI-first customer service platform with agents
No salary listed
Berlin, Germany
Hybrid
At least three days per week in the office required.
Bachelor's
 DevOps & Infrastructure (1) |
|---|
Get referred to Fin
See people who can refer or advise you
- At least 5 years of software engineering experience with a strong track record of shipping high-quality products or platforms.
- A degree in Computer Science, Computer Engineering, or a related field, or equivalent experience with very strong fundamentals.
- Hands-on experience with model training, especially transformers and large language models; model inference at scale, especially transformers and large language models; or low-level GPU work such as writing CUDA or Triton kernels.
- Comfort working in production environments at meaningful traffic, data, or organizational scale.
- Ability to communicate clearly and explain complex technical topics to different audiences.
- Strong technical fundamentals and willingness to invest in personal development.
- Deep knowledge of at least one programming language, such as Python, Ruby, Java, or Go, with the ability to write clean, reliable code and learn new technology stacks quickly.
- Implement and scale training pipelines for large transformer and large language model models, from data ingestion and preprocessing through distributed training and evaluation.
- Build and optimize inference services that deliver low-latency, high-reliability customer experiences, including autoscaling, routing, and fallbacks.
- Tune GPU-level kernels, improve utilization, and identify bottlenecks across the training and inference stack.
- Collaborate with machine learning scientists to implement cutting-edge training and inference methods and bring them to production.
- Participate actively in hiring, mentoring, and developing other engineers on the team.
- Raise technical standards, reliability, and operational excellence across the AI platform.
- Experience at AI-native companies that train or run inference for their own models, such as modern AI labs or AI-native product companies.
- Experience running training or inference workloads on Kubernetes.
- Experience with AWS or other major cloud providers.
- Production experience with Python in machine learning or infrastructure contexts.
- Demonstrated passion for technology through personal projects, open source, meetups, or publishing content about work and learnings.
Intercom offers an AI-first customer service platform centered on Fin AI Agent, a human-quality AI agent that can work with any Helpdesk, and an Intercom Customer Service Platform. The system runs on a single AI core and two main components to help businesses improve customer experiences, boost operational efficiency, and scale support as they grow. How it works: Fin AI Agent handles customer inquiries with human-like accuracy and can integrate with existing helpdesk tools, while the Intercom platform provides the surrounding customer service environment and workflows. What makes it different: it combines a unified AI core with a platform that works with any helpdesk, delivering consistent AI-powered support across channels and is trusted by thousands of support leaders. What is the goal: to help businesses deliver remarkable, efficient, and scalable customer service.
Company Size
1,001-5,000
Company Stage
Acquired
Total Funding
$490.8M
Headquarters
San Francisco, California
Founded
2011
Get referred to Fin
See people who can refer or advise you
Simplify's Take
What believers are saying
- Fin Voice 2 improved resolution rates 24.5% and cut latency 0.43 seconds.
- Fin claims 30,000 customers, $400 million ARR, and 2 million weekly interactions.
- Salesforce will extend Fin into smaller businesses, expanding channel reach immediately after close.
What critics are saying
- Salesforce’s $3.6 billion acquisition closes only in fiscal Q4 2027, delaying certainty.
- Agentforce overlap invites product cannibalization and forces Fin differentiation against Salesforce’s distribution.
- Competitors like Zendesk AI and Sierra attack Fin’s midmarket wedge on deployment speed.
What makes Fin unique
- Apex Flash and Apex 1.0 are purpose-built for customer-service resolution, not general chat.
- Fin Voice 2 handles refunds, identity checks, and workflows directly inside calls.
- Salesforce’s June 15, 2026 deal validates Fin’s product depth and midmarket reach.
Help us improve and share your feedback! Did you find this helpful?
Benefits
We reward our people - All full-time Intercom employees are offered a rewards program that includes competitive base salary, bonus, equity, and benefits.
We love to learn - There are so many opportunities here to learn and build a career. We support employees' growth by offering a wide range of continuing education options, including core skills, management training, and guided learning.
We take employee wellbeing seriously - We offer comprehensive healthcare coverage, including medical, dental, and vision, as well as employee assistance programs and mental health resources. We also offer flexible time off, significant paid family leave, and more.
The world has changed. So has the way we work - The majority of our employees say they prefer a hybrid model of work—some in-office time, some WFH—so that's our current approach. No matter what, though, we're committed to listening to our team.
Growth & Insights and Company News
6 month growth
↑ 40%1 year growth
↑ 38%2 year growth
↑ 39%Chatbase reaches $10M ARR as a profitable challenger to Sierra, Decagon, and Fin. * Jun 17, 2026 SAN FRANCISCO - June 16, 2026 -Chatbase, the AI customer support platform used by enterprise teams across more than 80 countries, has reached $10 million in annual recurring revenue. The company has never raised outside funding. Enterprise AI support is moving away from long, services-heavy deployments and toward platforms that prove value quickly, integrate into existing systems, and scale across channels without forcing companies into a single vendor stack. Chatbase has grown into that shift entirely on the strength of its product, without investor capital to accelerate adoption or subsidize pricing. Chatbase is already serving the kinds of global brands that define the enterprise AI support market, including IHG, Miele, National Grid, and Noon. These are not startup use cases. They are proof that enterprise teams are choosing a leaner, product-led platform over larger, more heavily funded vendors. The company is SOC 2 Type II certified, GDPR and HIPAA compliant, and runs on 35 AI models from seven providers. The market Chatbase competes in is now defined by heavily capitalized AI customer experience platforms, including Sierra, Decagon, and Fin, the company formerly known as Intercom. Those platforms have raised significant venture capital and are building toward outcomes that require broad market ownership. Chatbase's growth demonstrates something different: that enterprise customers are actively choosing platforms that deploy quickly, integrate deeply, and improve the economics of support without adding operational complexity or vendor dependency. Chatbase's lack of outside funding is not a constraint. It is part of the enterprise value proposition. A platform that does not need to own the entire market to succeed has different incentives than one that does. There is no forced lock-in, no pricing pressure from a venture outcome that requires aggressive expansion, and no conflict between what investors need and what customers need. The company reached $1 million ARR in 117 days from launch, entirely through organic channels, with no paid marketing. The platform now handles customer support across chat, voice, email, and messaging through a single agent configuration. When Chatbase Voice launched in May 2026, it extended the same knowledge base, trained actions, and escalation logic that powers the chat agent to inbound phone calls via Twilio, with support for 95 languages and no additional vendor. For enterprise teams managing high inbound volume across time zones, that means one system, one setup, and no operational overhead from running separate tools for separate channels. "10 million is just a number, but I think it signifies that we were able to do this basically without spending any money in the early days," said Yasser Elsaid, Founder and CEO of Chatbase. "Now I think this is a turning point where we can operate as a venture-backed company because we've had this revenue from the last two years. If we operate like a venture-backed company, we can get to 100 million in the next few years. Maybe two." Chatbase is now accelerating its enterprise expansion, investing in a dedicated sales function and targeting the mid-market and enterprise accounts that represent the bulk of its addressable market. With 26 employees, a profitable operating model, and an omnichannel product already running at enterprise scale, the company enters that push from a position that most of its competitors, despite their funding, have not yet reached. About Chatbase Chatbase is an AI agent platform for customer support used by more than 10,000 businesses worldwide, including IHG, Miele, National Grid, and Noon. The platform lets companies build, train, and deploy AI agents that resolve customer issues across chat, voice, email, and messaging by connecting to their knowledge base, internal tools, and CRM systems. Chatbase supports 35 AI models from 7 providers, custom actions, human escalation, responses in 95 languages, and native integrations with Zendesk, Salesforce, Intercom, HubSpot, Stripe, Shopify, Slack, WhatsApp, and more. Media Contact Zeyad Genena Chatbase Inc. 2261 Market Street, Suite 85690 San Francisco, CA 94114 Media Contact Company Name: Chatbase Contact Person: Zeyad Genena Country: United States
Salesforce buys Fin, targets midmarket. The $3.6 billion acquisition of the CX-focused AI agent platform will strengthen Salesforce's potential move into the midmarket. June 16, 2026 No Jitter Brief: * Salesforce will acquire Fin, a customer service-focused AI agent company for $3.6 billion. Fin was formerly Intercom, the 1,400-employee strong, 15-year-old company, rebranded in May 2026. * Fin's core offering is its AI Agent, powered by a proprietary AI model "purpose-built for customer support," which resolves customer queries end-to-end, across multiple channels including live chat, email, WhatsApp, SMS, phone and Slack. In June, Fin debuted Operator, an agent for customer operations "that spots problems, finds opportunities to improve and acts on them," per the company. * According to Sheila McGee-Smith, founder and principal analyst at McGee-Smith Analytics, the Fin purchase will help boost Salesforce's position with small and medium companies and inject some needed verve into the Agentforce platform. She told No Jitter, "Fin's March 2026 announcement of Apex 1.0, a CX model suite purpose-built or customer service, and the June announcement of Operator, built on Apex, looks to me like something that can bring start-up energy to the Agentforce program." No Jitter Insight: CCaaS vendors like Zoom, Webex, RingCentral, 8x8, GoTo, Zendesk and others, all have offerings targeted at mid-market firms. The mid-market includes firms with annual revenues between $10 million and $1 billion, with employee counts anywhere from 50 to 1,000. As shown below, 4 in 10 midmarket firms said they intend to moderately increase total spending (between 5 to 15%) on AI-powered CX technology, per a recent Omdia study. Two in 10 plan to significantly increase (15% or more) their total spending. Additionally, 68% of midmarket firms said that AI will be either foundational to most CX initiatives (34%) or very important, but not the primary driver (34%). Estimates vary, but research from the National Center for the Middle Market found that there were "nearly 200,000 U.S. middle market businesses that represent one-third of private sector GDP, employing approximately 48 million people" at year-end 2025. Fin's purpose-built customer service AI "will help Salesforce accelerate execution and improve service outcomes faster, especially in the mid-market where Fin (Intercom) has strategically been focused," Mila D'Antonio, a principal analyst on Omdia's Customer Engagement team, told No Jitter in an email. "Fin's AI capabilities can deepen Salesforce's market penetration in the mid-market by gaining access to a strong customer base to help it compete more aggressively in this rapidly evolving market. Ultimately it will strengthen Salesforce's position." Salesforce entered territory adjacent to its core CRM business when it launched Agentforce Contact Center in March. Salesforce spent 15 months building that product and "hired experts in telephony and contact centers from leading vendors," wrote McGee-Smith in her No Jitter article discussing the news. Salesforce described ACC as the "only AI contact center solution that unifies voice, digital channels, CRM data and AI agents natively in a single system." Fin's Apex was post-trained "using years of proprietary customer service data accumulated through Fin," per VentureBeat. Post-training is a process that makes the foundation model more useful in particular scenarios, like customer service. Fin currently resolves 2 million customer interactions per week and boasts resolution rates "averaging 76% across 12,000+ customers, with many seeing over 85%," per the company's site. Along with adding Fin's capabilities to Agentforce, Salesforce will also add Fin's technical AI team and its 30,000-strong global customer base to Salesforce, per the release. "Salesforce is pivoting into a new area where they do not have experience or expertise. I understand their need to do so, but they are going to tarnish their brand if they attempt to build this on their own without all the years of evolution which other companies in this space have completed," Dave Michels, founder and principal analyst at TalkingPointz told No Jitter in an email. "Making some smart acquisitions is the most logical thing to do." The transaction is expected to close in the fourth quarter of Salesforce's fiscal year 2027. Disclosure: Informa TechTarget, the publisher behind No Jitter, also owns Omdia. Informa TechTarget has no influence over No Jitter's coverage. Senior Editor Matt Vartabedian is the senior editor for No Jitter, a role he has held since March 2023. Matt covers generative and agentic AI as it pertains to enterprise communications - i.e., unified communications and collaboration (UC&C) and the contact center. He also covers related areas, including customer experience (CX) and customer relationship management (CRM). Matt began his B2B journalism career in the late 1990s writing for several call center and telecommunications print magazines. He then spent two decades as a cellular industry analyst, where he authored market reports, articles, presentations and opinion pieces grounded in significant primary and secondary research. You can reach Matt at [email protected] and on LinkedIn. Want more No Jitter stories in your Google search results?
Hot take: Salesforce acquires fin and leaves Constellation Research, Inc. reading between the agentic lines. June 16, 2026 There is an optimist's path and a pessimist's path to absorbing the news that Salesforce has entered into an agreement to acquire Fin, the customer-facing AI agent that until May 2026 had been known as Intercom. But first, here is what Constellation Research, Inc. know about the deal. Salesforce has agreed to acquire fin for a reported $3.6 Billion. The deal, once it clears all customary regulatory reviews and approvals, is expected to close in Salesforce's Q4 fiscal 2027. According to Fin's website, Fin resolves over 2 million conversations weekly, has doubled in growth year over year, with a self-reported $400M in annual recurring revenue. The company, despite being based in San Francisco, is widely hailed as a shining example of Irish tech innovation success, with Irish news outlets heralding the acquisition announcement as the largest-ever acquisition of an Irish tech company. What Constellation Research, Inc. Know About Fin: The best place to start about why this deal is interesting is not with Fin's expertise in CX-focused agents serving sales, service, commerce and marketing engagement needs, but rather with Fin's proprietary frontier model, Fin Apex 1.0. Billed as a purpose-built model for customer service, Apex 1.0 claims to be faster, more accurate and effective than other less focused models in the market, with 65% fewer hallucinations and 2.8% higher resolution rates than competitors like OpenAI's GPT 5.4 and Anthropic's Sonnet 4.6. Fin has developed a suite of models with the center point of customer answers formed by Apex 1.0. However, a host of other models address everything from summarization and translation, escalation, content retrieval and "reranker" for optimized knowledge and LLM answer availability. The Fin agentic vision is that there is a single customer agent, as opposed to individual functionally-bound agents. This single customer agent is trained to address the entirety of a customer's journey, managing leads through to transaction and success. The Optimists' View: AgentForce was arguably born from an Intercom competitor, Airkit. While that conversational play became the first phase of what Salesforce would eventually articulate as an agentic enterprise powered by a hybrid workforce of humans enhanced and assisted by agents. Salesforce CEO Marc Benioff is not content with the resolutions and conversational completions Agentforce can deliver today...and envisioning more from Salesforce is a quintessentially Benioff brand. Fin Apex will join xLAM (a large action model built to address complex tasks) and xGen-Sales (a model trained and designed to address autonomous actions in sales), joining the Salesforce AI Research group's suite of models and the expansive array of connectors and integrations Salesforce has forged as part of its larger bring-your-own-model approach and Einstein Trust Layer. As Salesforce has worked to deliver a comprehensive AI and data infrastructure for organizations of all sizes to leverage, in the press announcement for the deal, Benioff specifically pointed to Fin being an opportunity to rapidly ramp up adoption for even the smallest of businesses. While Agentforce has always been billed as easy and fast to build an agent, connecting those agents to business goals and outcomes can be difficult to envision for organizations working to do more with less. Salesforce is envisioning Agentforce with Fin as a supercharged high-speed onramp to agentification. But look deeper into the Benioff crystal-ball of innovation and you start to see a dismantling of the very functional-cloud segmented business that Salesforce helped pioneer. In fact, when you take this deal layered on top of the recent announcement of Salesforce's agreement to acquire headless content powerhouse, Contentful, you start to see a massive reshuffling of the deck where clouds start to thin in favor of focused customer actions. It begs the question if Salesforce is itself starting to rethink the platforms powering Engagement, Growth and Success as curated agentic processes as opposed to the more traditional functionally rigid Sales, Service, Marketing and Commerce clouds. Rather than functional leaders owning processes and tasks, new teams armed with agents and data can tackle strategies like customer experience as an enterprisewide team sport, rather than fiefdoms battling for ownership, credit and control. It would restructure Salesforce as an offering and as a business, but it needs the core customer agent to quickly wrap around that Customer 360 vision to work. The Pessimist's View: The pessimist's biggest question will be "why?" More specifically, why does Salesforce need more agentic capabilities? Is this a sign that AgentForce isn't meeting expectations? Between the Contentful and the Fin acquisition, it can feel like Salesforce is filling gaps in its big vision of Headless 360 and Agentforce. Critics and competitors have been quick to float the rumor that this buying spree is a return to the M&A patterns of the past...a past that supposedly sparked activist investors to demand more financial rigor. Salesforce is no stranger to the criticism that their big main stage vision keynotes take way too much time to manifest. Salesforce is also dealing with the very real issue of customer hesitation to modernize applications while they simultaneously demand access to all the AI goodness of public cloud and the lightning platform. It will take more than vision and charisma to break decades of functional dysfunction. Sales, marketing and service have fought for generations, honing finger pointing, hoarding assets and fortifying rigid silo walls that serve as insular tombs where customer data can go to die. To achieve the idea of a single customer agent, perceived ownership of the customer will need to be abandoned for the good of the customer. For too many organizations, it is too much change and far too much to ask. Final Thoughts: This won't be the last agentic pickup Constellation Research, Inc. see Salesforce make in their fiscal-27 year. The real question will be where Salesforce sees opportunity to keep expanding the Agentforce net across platform, data, security or applications. There is always plenty of vision, innovation and bold creativity at Salesforce. But as the market continues to rumble with hard questions around overarching value, outcomes and what the real cost of AI should be, customers want to hear what is capable today, and less about what a future-forward agentic fever dream could look like down the road. Salesforce is being asked to once more to emerge as a leader, running at multiple speeds to meet customers where they stand today while continuing to foster big AI aspirations. Vice President & Principal Analyst Constellation Research About Liz Miller: Liz Miller is Vice President and Principal Analyst at Constellation, focused on the org-wide team sport known as customer experience. While covering CX as an enterprise strategy, Miller spends time zeroing in on the functional demands of Marketing and Service and the evolving role of the Chief Marketing Officer, the rise of the Chief Experience Officer, the evolution of customer engagement, and the rising requirement for a new security posture that accounts for the threat to brand trust in this age of AI. With over 30 years of marketing experience, Miller offers strategic guidance on the leadership, business transformation, and technology requirements to deliver on today's CX strategies. She has worked with global marketing organizations to transform everything from... Published. June 16, 2026
Salesforce to buy AI agent platform Fin for about $3.6 billion. 15 Jun 2026 08:25PM (Updated: 15 Jun 2026 08:39PM) Add CNA as a trusted source to help Google better understand and surface our content in search results. June 15: Salesforce said on Monday it had signed an agreement to buy autonomous AI agent platform Fin for about $3.6 billion. The deal strengthens Salesforce's Agentforce platform, as technology firms compete to roll out usage-based autonomous digital workers across enterprises. Following completion, Salesforce and Fin will expand options for deploying AI agents in customer service, allowing customers to incorporate them with existing systems, the companies said. Fin makes an AI customer agent that handles support questions across live chat, email, WhatsApp, SMS, phone and Slack. The deal is expected to close in the fourth quarter of Salesforce's fiscal year 2027, and no change is expected to the company's forecast and capital return program.
Salesforce acquires Fin for $3.6 billion. Published June 15, 2026 Salesforce said it is acquiring Fin, formerly known as Intercom, for $3.6 billion in a deal that brings customer service AI agents and a proprietary AI model called Apex to Agentforce. At first glance, Fin's focus on AI agents and a platform aimed at resolving customer support looks a bit redundant. Fin also has offerings for service, sales and e-commerce as well as integrations with Freshdesk, Salesforce and HubSpot. Many of Fin's capabilities overlap with what Salesforce's Agentforce does across multiple channels. In a statement, Salesforce CEO Marc Benioff said "Fin brings proven agent technology" that will complement Agentforce's service agents. According to the companies, Salesforce and Fin will combine to give customers more ways to deploy AI agents across customer service. Salesforce noted that Fin will extend its reach into the SMB market. Fin will be able to scale its business as part of Salesforce. Salesforce, which recently acquired Contentful, said the deal will close in its fiscal fourth quarter. Key things to know about Fin: * The company's Customer Agent is one agent that works across qualifying leads, closing sales and resolving support issues. The agent can also manage e-commerce and switch roles as needed. * Fin's platform has roles, customer facing agents, agent operations, tools for governance and control, and purpose-built models. * Fin's flagship model is Apex 1.0. * The big idea from Fin is to use one agent to work across an entire customer journey. Here are a few screens of Fin's platform. Editor in Chief of Constellation Insights Constellation Research Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context... Results. Insight News June 14, 2026 Next-Generation Customer Experience Starbucks CEO Brian Niccol is betting the coffee company's returns on AI investment will come from outcomes that aren't noticed by customers that much. The company is focusing on s... Larry Dignan Insight News June 13, 2026 Data to Decisions The US government cited national security concerns and suspended access to any foreign national using Fable 5 and Mythos 5 inside or outside the US. Anthropic then moved to disable... Larry Dignan Insight News June 12, 2026 Innovation & Product-led Growth SpaceX's financial results are driven by SpaceX and Starlink with the AI unit, including X and xAI being a drag... Larry Dignan Insight News June 11, 2026 Future of Work Adobe's second quarter results were well ahead of estimates as it continues to see demand for its AI tools. The company also raised its outlook... Larry Dignan Insight News June 10, 2026 Data to Decisions Oracle is moving toward outcome-based pricing as a model to complement its existing enterprise contracts. The big question is whether Oracle and its customers can deliver the outco... Larry Dignan Insight News June 10, 2026 Data to Decisions Oracle reported better-than-expected fourth quarter results, delivered 47% in cloud revenue growth compared to a year ago and 93% growth in infrastructure as a service... Larry Dignan Published. June 15, 2026 Insight News June 15, 2026 Next-Generation Customer Experience Salesforce said it is acquiring Fin, formerly known as Intercom, for $3.6 billion in a deal that brings customer service AI agents and a proprietary AI model called Apex to Agentfo... Larry Dignan Insight News June 14, 2026 Next-Generation Customer Experience Starbucks CEO Brian Niccol is betting the coffee company's returns on AI investment will come from outcomes that aren't noticed by customers that much. The company is focusing on s... Larry Dignan Insight News June 11, 2026 Future of Work Adobe's second quarter results were well ahead of estimates as it continues to see demand for its AI tools. The company also raised its outlook... Larry Dignan