Unsiloed AI (YC F25)’s cover photo
Unsiloed AI (YC F25)

Unsiloed AI (YC F25)

Software Development

San Francisco, California (CA) 5,157 followers

AI Agents for Unstructured Financial Data

About us

We are a Y Combinator backed startup building AI agents that simplify working with unstructured financial data, turning complex documents into structured, actionable insights. We make working with unstructured data as seamless as running SQL queries. Our proprietary vision-language model ensures accurate data extraction, seamless ETL integration, and compliance-ready workflows for financial institutions

Website
https://www.unsiloed-ai.com
Industry
Software Development
Company size
2-10 employees
Headquarters
San Francisco, California (CA)
Type
Privately Held
Founded
2024
Specialties
AI, Unstructured Data, Financial Workflows, VLMs, and Extraction

Locations

Employees at Unsiloed AI (YC F25)

Updates

  • Unsiloed AI (YC F25) reposted this

    View profile for Hunter Wittenborn

    Programmer, DevOps, FOSS, and heavy user of automating all the things™.

    My team (Slava Iudenko, Siddharth Singh, Nick Yeung) just wrapped up our 24-hour MVP at the Y Combinator Agent Jam hackathon this past weekend! This event involved building an agent and utilizing it in creative ways. We were able to hack together Perp, an assistant designed to give teams full context-awareness for their meetings: - Ingests context from documents, transcriptions, and more - Provides real-time awareness for live meeting participants (especially helpful when you show up late 😉) - Ensures the entire team stays in the loop after the meeting ends with meeting summaries and action items While we didn't win the overall competition, our team was able to snag a victory through a sub-track hosted by Unsiloed AI (YC F25)! Beyond the project, I was absolutely blown away by the quality of people in California! From the second I stepped on my plane, I was immediately connecting with other founders, and it was incredibly inspiring to be surrounded by such a vast amount of highly-driven, like-minded people. A huge shout-out to our judges, leaders, and mentors, including Andrew Miklas, Aman Mishra, Karim Rahme, Tobias Herber, Joseph Jacob, and so many others! (I'd name them all if I weren't forgetting names from lack of sleep 😅) This event is just the beginning of what all of these talented individuals are building. Now it's time to keep moving forward, keep making progress, and make something people want!

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
  • Unsiloed AI (YC F25) reposted this

    View profile for James Wilson

    CSEN @ SCU | Building & Scaling Infrastructure for 300k+ MAU

    This past weekend was one of the most inspiring weekends I’ve had in a long time. I applied and was admitted to Y Combinator and Metorial (YC F25)’s Agent Jam hackathon hosted at the Y Combinator hq in downtown San Francisco. The event brought together an incredible group of founders, engineers, and builders all experimenting with new ways to push the limits of AI. Over the weekend, I built an AI handwriting model that takes text and generates handwriting in different styles. I trained, tested, and implemented the model. I also implemented Captain (YC F25) into our project, a system that enables infinite context windows for LLMs, letting our project recall and reason across unlimited user uploaded information. We didn’t walk away with a prize, but I walked away with something better. New ideas, new connections, and a deeper appreciation for how much can be accomplished when you’re surrounded by passionate people who love to build. Huge thanks to Y Combinator, Metorial (YC F25), Coval, Unsiloed AI (YC F25), Captain (YC F25), and all the other sponsors for hosting such an inspiring event and to everyone I met for making it such a memorable experience.

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
  • Unsiloed AI (YC F25) reposted this

    View profile for Mian M.

    Product Designer/Engineer

    Last week, I took on my first 24-hour hackathon at Y Combinator’s Agent Jam ‘25, held by Metorial (YC F25) and the experience was next-level with my teammate Zin Khant The hackathon brought together some of the most talented developers, students, and engineers in San Francisco all geared up to push the limits of what AI agents can do, and even compete for a shot at a guaranteed YC interview. Metorial transformed the Y Combinator office into a builder’s playground, complete with tracks supported by batchmates like Captain (YC F25),Unsiloed AI (YC F25), Nivara, Audrion, Bear, and Mantle. Sponsors including Natural, Vercel, OpenAI, Anthropic, Rippling, and Violet Ventures made the event possible. It was inspiring to see so many future founders and AI enthusiasts collaborating under one roof. For the main event, my team built an  Agentic Survey Tool using Metorial, powered by Exa search MCPs. The ability to deploy from such a massive library with just a couple clicks was a game changer! Meeting talented builders from all over, solving complex problems, and sharing adrenaline fueled moments made the whole hackathon unforgettable. I came away motivated and ready for whatever comes next.

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
  • Unsiloed AI (YC F25) reposted this

    View profile for Slava Iudenko

    Student Innovation Assistant @ UW Tech Exploration Lab | Software Engineering, Machine Learning

    Just spent over 24 hours straight caffeinating and building an MVP at the Y Combinator HQ. For the YC Agent Jam Hackathon this weekend, Siddharth Singh, Nick Yeung, Hunter Wittenborn and I hacked up a pretty cool AI Meeting Assistant and earned a win in the Unsiloed AI (YC F25) track. Huge thank you to the judges, including Andrew Miklas, as well as Tobias Herber and Karim Rahme at Metorial (YC F25) for putting such an amazing event together. I learned a lot, met many amazing founders, and was absolutely blown away by the energy. And yeah, the third photo is us hitting a few mandatory 6-7's at YC in honor of GPT-6 being renamed to GPT-6-7. In case you were wondering.

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
  • Most accurate APIs to parse multimodal unstructured data and make them ready for your LLMs and AI Agents.

    View organization page for Y Combinator

    1,506,500 followers

    Unsiloed AI (YC F25) builds the most accurate APIs for parsing multimodal unstructured data like PDFs, PPTs, tables, charts, and images into structured Markdown and JSON that downstream LLMs and AI agents can reliably use.
 Today, nearly 80% of enterprise data lives in unstructured formats from medical records and contracts to financial statements, reports, and insurance claims. Going forward, there will be more AI agents reading from and writing to documents than humans, especially across verticals like finance, legal, and healthcare that still run more on documents than databases. AI teams spend months building document-ingestion pipelines that constantly break with layout changes or inconsistent formatting. From the tons of open-source solutions out there, it’s still tough to achieve production-grade accuracy on even mildly complex documents. Traditional OCRs are static and fail with changing layouts. LLMs, though great at comprehension, struggle with deterministic extraction, making them unreliable for accuracy-sensitive verticals like finance. As a result, AI teams end up becoming document AI companies, reinventing the wheel.
 Unsiloed AI solves this by using their proprietary Vision Language Models with a dual-stream architecture that preserves the structure and hierarchy apart from the content, novel pre-processing and segmentation, and domain-specific decoders that structure outputs into JSON or Markdown with confidence scoring and reinforcement learning loops. It can run in fully on-prem and air-gapped environments, and already processes millions of pages each week for Fortune 150 banks, public insurance companies, and fast-moving startups across finance, legal, and healthcare.
 Unsiloed was founded by Aman Mishra and Adnan Abbas, batchmates from IIT Kharagpur. Aman was a Quant Developer at a hedge fund, building ultra-low-latency trading systems moving billions across global exchanges, and later the first engineer at a startup where he built AI products used by Goldman Sachs and Charles Schwab, where he faced this problem firsthand. Adnan worked on autonomous navigation at Mercedes and was a researcher at MIT, where he developed multimodal models deployed at a Fortune 10 company. Their goal is simple: make accurate multimodal data ingestion as easy as a single API call, making documents as accessible for AI agents and LLMs as the data sitting in relational databases. Try it out at unsiloed.ai. Congrats to the team on the launch! 🚀 https://lnkd.in/gfj59z-4

  • From S3 to semantic search in 4 steps: Fetch → Parse → Embed → Store Our Vision API turns static docs into a searchable knowledge base Our latest blog walks through a complete pipeline: Fetch from S3 → Parse with Vision API → Embed with OpenAI → Store in Pinecone. The result? Fast, meaning-based search across millions of documents that is ideal for compliance, contracts, and enterprise knowledge management. Read the full guide (with code + Colab notebook): Blog: https://lnkd.in/gubvVQMm Notebook: https://lnkd.in/g4mVcJPj Visit unsiloed.ai for more

    • No alternative text description for this image

Similar pages

Browse jobs

Funding