Jarmin
Autonomous ML engineer agent via chat for production ML pipelines
Website: https://www.jarmin.ai/
Cover Block
PUBLIC
The company is a recent Y Combinator entrant aiming to automate the work of a machine learning engineer through a conversational interface.
| Attribute | Details |
|---|---|
| Name | Jarmin |
| Tagline | Autonomous ML engineer agent via chat for production ML pipelines |
| Headquarters | San Francisco, CA, USA |
| Founded | 2025 |
| Stage | Seed |
| Business Model | B2B |
| Industry | Deeptech |
| Technology | AI / Machine Learning |
| Geography | North America |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding Label | Undisclosed |
Links
PUBLIC
- Website: https://www.jarmin.ai/
- LinkedIn: https://www.linkedin.com/company/jarmin-ai/
- X / Twitter: https://x.com/jarmin_ai
Executive Summary
PUBLIC
Jarmin is an early-stage bet that the bottleneck for applied machine learning is shifting from model access to engineering execution, and its proposition is to act as an autonomous ML engineer delivered via chat. Founded in 2025 and currently in Y Combinator's Fall cohort, the company targets mid-market teams with data science backlogs but limited in-house ML engineering talent [Y Combinator, 2025].
Its core product is described as a system of proprietary agents that connects to a client's data, runs experiments, builds models, and manages deployment pipelines, ostensibly handling the hundreds of micro-decisions that typically require a human engineer [jarmin.ai, 2025]. The initial go-to-market is founder-led, selling scoped engagements rather than a self-serve product, with an implied average contract value between $100k and $150k [FYI Combinator, 2025].
The founding team includes Chalmers Brown, a former Meta engineer and CTO of personal finance platform Due, alongside Shirsendu Halder, who holds a master's from Carnegie Mellon's Robotics Institute, and Zack Leman [LinkedIn, 2026] [Carnegie Mellon University, 2026] [Due, 2026]. Capitalization is not publicly disclosed beyond the Y Combinator affiliation, and no named customers or detailed case studies have been published.
For investors, the next 12-18 months will be defined by the company's ability to convert its Y Combinator momentum into a handful of referenceable enterprise deployments, validate its pricing and service model at scale, and begin the technical transition from a service wrapper to a more productized, scalable platform.
Data Accuracy: YELLOW -- Core company claims are sourced from its website and Y Combinator page; founder backgrounds are partially corroborated by LinkedIn and university records. Market sizing and business model details are inferred from a single third-party analysis.
Taxonomy Snapshot
| Axis | Classification |
|---|---|
| Stage | Seed |
| Business Model | B2B |
| Industry / Vertical | Deeptech |
| Technology Type | AI / Machine Learning |
| Geography | North America |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding | Undisclosed |
Company Overview
PUBLIC
Jarmin is a very early-stage venture, founded in 2025 and headquartered in San Francisco, California [Crunchbase, 2025]. The company's primary public milestone is its admission into the Y Combinator accelerator program as part of the Fall 2025 batch, which serves as the current anchor for its timeline [Y Combinator, 2025]. No details regarding a pre-YC founding story, incorporation date, or specific legal entity are available in public registries or coverage.
The founding team consists of three co-founders: Chalmers Brown, Shirsendu Halder, and Zack Leman [jarmin.ai, 2025]. Public profiles for the founders show educational backgrounds at Rutgers University, Carnegie Mellon University's Robotics Institute, and Bowdoin College, respectively [Chalmers Brown LinkedIn, 2026] [ri.cmu.edu, 2026] [Zack Leman LinkedIn, 2026]. Brown is also listed as a former co-founder and CTO at a separate fintech company, Due [Due, 2026]. The company claims its broader team includes individuals with prior experience at Meta, Apple, AWS, Lockheed Martin, and JPMorganChase, though these affiliations are not individually verified [jarmin.ai, 2025].
Beyond the Y Combinator launch, no subsequent product launches, customer announcements, or funding rounds have been publicly disclosed. The company's operational model is described as founder-led, scoped engagements, indicating a services-oriented go-to-market in its initial phase [FYI Combinator, 2025].
Data Accuracy: YELLOW -- Founding year and YC participation are confirmed. Founder names and educational details are partially corroborated by LinkedIn and university records, but prior work experience and team composition claims are sourced solely from the company.
Product and Technology
MIXED Jarmin's core proposition is an autonomous agent that executes the end-to-end machine learning lifecycle through a conversational interface. A user describes an ML task, and the system, according to its Y Combinator launch description, connects to data sources, runs experiments, builds models and pipelines, and deploys to production, incorporating optional human approval gates [Y Combinator, 2025]. The company's website frames this as a "system of proprietary agents, models, automations that each handle portions of the ML lifecycle," aiming to automate the hundreds of micro-decisions typically made by human ML engineers [jarmin.ai, 2025].
The underlying technology stack is not publicly detailed. The product is not a self-serve platform; go-to-market is conducted through "scoped, founder-led engagements" where the Jarmin team presumably operates the agent on behalf of a client to deliver a specific model or pipeline outcome [FYI Combinator, 2025]. This service-model approach suggests the initial product is tightly integrated with the founding team's operational expertise.
Data Accuracy: YELLOW -- Product claims are sourced from the company's YC page and website; the service-model detail is from a third-party analysis. Technical architecture and stack details are not confirmed.
Market Research
PUBLIC The market for tools that accelerate the productionization of machine learning is expanding as the gap between experimental models and deployed systems becomes a primary constraint on enterprise AI adoption.
Jarmin's initial market sizing is derived from a single source, which frames the opportunity around service engagements for a specific customer segment. According to an analysis from FYI Combinator, the company's implied total addressable market (TAM) is estimated at $1.0 billion to $3.0 billion [FYI Combinator, 2025]. This figure is calculated on the assumption of selling engagements priced between $100,000 and $150,000 annually to a pool of 10,000 to 20,000 mid-market teams that are described as having existing ML backlogs. This is a service-oriented SAM (serviceable addressable market) estimate, not a product TAM, and it is contingent on the company's ability to consistently close and deliver on these founder-led contracts.
Demand drivers for this category are well-documented in broader industry research, though not specifically cited for Jarmin. The persistent shortage of specialized ML engineering talent is a primary catalyst, creating a backlog of projects within data science teams [analogous market, Gartner, 2024]. A secondary driver is the increasing pressure on product teams to integrate predictive features, such as recommendation or churn models, which necessitates moving beyond prototype notebooks to stable, monitored pipelines. These tailwinds suggest a growing willingness among mid-market companies to pay for outcomes that bypass internal hiring and development bottlenecks.
Key adjacent markets that function as substitutes or expansion vectors include the broader MLOps platform space, low-code AI application builders, and managed data science consultancies. Jarmin's chat-driven, autonomous agent pitch positions it between a full-service consultancy and a self-serve software platform. Regulatory or macro forces are not a primary focus in the available materials, though the general trend toward AI governance and model risk management could eventually influence the feature requirements for any production deployment tool.
Service SAM (Mid-Market) | 1.5 | $B
The single-point market estimate, while logically constructed, highlights the early and unproven nature of Jarmin's commercial model. The entire business case rests on converting a theoretical population of underserved teams into paying clients at a premium price point.
Data Accuracy: YELLOW -- Market sizing is from a single third-party analysis (FYI Combinator) and is not corroborated by independent industry reports.
Competitive Landscape
MIXED Jarmin enters a crowded automation market not with a tool for engineers, but with a service that aims to replace them for a specific customer segment.
The competitive map for Jarmin is defined by three distinct layers of alternatives, each serving a different need. At the incumbent level, established ML platform vendors like Databricks, AWS SageMaker, and Google Vertex AI provide the foundational tooling and infrastructure for teams that already have ML engineering talent. These are not direct competitors but represent the entrenched, do-it-yourself paradigm. The challenger layer consists of newer companies focused on automating parts of the ML workflow, such as Hasty.ai for computer vision or Arize AI for model monitoring. These point solutions automate specific tasks but still require an integrated team. The most direct adjacent substitutes are the consulting firms and agencies, from large system integrators to boutique ML shops, which provide the human expertise Jarmin seeks to productize. Jarmin's positioning attempts to carve a path between these layers by offering a productized, chat-driven service that promises end-to-end automation without the need for internal specialists or the high cost and project risk of traditional consulting [FYI Combinator, 2025].
Jarmin's current edge is its founder-led, service-wrapped go-to-market, which acts as both a wedge and a moat. By selling scoped engagements directly, the founders can deeply understand customer pain points, iterate on the proprietary agent system in real-world conditions, and build a corpus of domain-specific playbooks. This hands-on model creates a feedback loop that pure software vendors lack and allows Jarmin to command a premium ($100k-$150k ACV) for outcomes, not just software access [FYI Combinator, 2025]. The durability of this edge, however, is perishable. It is contingent on the founders' bandwidth and the company's ability to systematically encode learnings from these engagements into its autonomous system. If the service model does not rapidly scale or the automation fails to reduce human involvement over time, the edge evaporates, leaving Jarmin as just another high-touch consultancy.
The company's most significant exposure lies in its narrow focus on the mid-market team with an ML backlog. This segment is also targeted by low-code/no-code platforms that empower data analysts, and by the expanding capabilities of the incumbent cloud platforms which are increasingly bundling automation features. A named competitor like DataRobot, which has pivoted toward enterprise automation, could use its existing sales footprint and brand recognition to move downmarket with a similar offering, undercutting Jarmin's founder-led sales motion. Furthermore, Jarmin does not own a proprietary data asset or a unique distribution channel; its entire value proposition is currently locked inside its yet-to-be-proven agentic system, which remains a black box to external observers [jarmin.ai, 2025].
Over the next 18 months, the most plausible competitive scenario is a bifurcation. If Jarmin successfully productizes its service learnings and launches a scalable, self-serve tier, it could become the "winner if automation reliability hits a threshold," capturing the long-tail of underserved teams and forcing incumbents to acquire or build competing agentic layers. The "loser if consulting margins compress" would be the boutique ML agencies, which would find their value proposition eroded by a cheaper, faster, automated alternative. However, if Jarmin remains stuck in a high-touch service model, it risks being outmaneuvered by a well-funded automation platform that decides to build a similar agentic layer on top of its existing infrastructure, leveraging a much larger installed base.
Data Accuracy: YELLOW -- Competitive analysis is inferred from company positioning and market structure; no direct competitor comparisons from independent sources.
Sources
PUBLIC
[Y Combinator, 2025] Jarmin: 24/7 Machine Learning Engineer employees | https://www.ycombinator.com/companies/jarmin
[jarmin.ai, 2025] Jarmin AI | https://www.jarmin.ai/
[Crunchbase, 2025] Jarmin - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/jarmin
[FYI Combinator, 2025] Jarmin | FYI Combinator | https://fyicombinator.com/company/jarmin
[LinkedIn, 2026] Chalmers Brown - Jarmin (YC F25) | https://www.linkedin.com/in/chalmersbrown/
[Carnegie Mellon University, 2026] Shirsendu Halder - Robotics Institute Carnegie Mellon University | https://www.ri.cmu.edu/ri-people/shirsendu-halder/
[Zack Leman LinkedIn, 2026] Zack Leman - Jarmin | LinkedIn | https://www.linkedin.com/in/zackleman/
[Due, 2026] Chalmers Brown, Former CTO and Author at Due | https://due.com/author/chalmers-brown/
[Chalmers Brown LinkedIn, 2026] Chalmers Brown - Jarmin (YC F25) | https://www.linkedin.com/in/chalmersbrown/
[ri.cmu.edu, 2026] Shirsendu Halder - Robotics Institute Carnegie Mellon University | https://www.ri.cmu.edu/ri-people/shirsendu-halder/
Articles about Jarmin
- Jarmin's Autonomous ML Agent Aims for the $100,000 Backlog — The YC-backed startup is selling scoped engagements to mid-market teams, betting its system of agents can handle the micro-decisions that stall production models.