Hacker News: diptanu

Building Scalable GitHub Runner Infrastructure on Sandboxes

diptanu — Mon, 03 Aug 2026 17:32:23 +0000

Article URL: https://www.tensorlake.ai/blog/build-your-own-ci-infrastructure

Comments URL: https://news.ycombinator.com/item?id=49158912

Points: 1

# Comments: 0

Firecracker disk snapshots in O(changed bytes), not O(disk size)

diptanu — Tue, 14 Jul 2026 17:46:05 +0000

Article URL: https://www.tensorlake.ai/blog/firecracker-disk-snapshots-o-changed-bytes

Comments URL: https://news.ycombinator.com/item?id=48910487

Points: 3

# Comments: 0

Zero-copy TLS ingress with kTLS and splice(2) for sandboxes

diptanu — Fri, 10 Jul 2026 15:46:38 +0000

Article URL: https://www.tensorlake.ai/blog/near-zero-overhead-sandbox-networking

Comments URL: https://news.ycombinator.com/item?id=48861533

Points: 1

# Comments: 0

Filter and Rank: Robust Multi-Cloud Sandbox Orchestration at Scale

diptanu — Wed, 13 May 2026 22:45:19 +0000

Article URL: https://www.tensorlake.ai/blog/multi-cloud-scheduling

Comments URL: https://news.ycombinator.com/item?id=48128581

Points: 1

# Comments: 0

Sandbox plumbing infrastructure for computer-use agents

diptanu — Fri, 01 May 2026 15:17:28 +0000

Article URL: https://www.tensorlake.ai/blog/building-sandboxes-for-computer-use

Comments URL: https://news.ycombinator.com/item?id=47975808

Points: 3

# Comments: 0

New comment by diptanu in "Show HN: Sub-millisecond VM sandboxes using CoW memory forking"

diptanu — Wed, 18 Mar 2026 01:45:08 +0000

The tricky part of doing this in production is cloning sandboxes across nodes. You would have to snapshot the resident memory, file system (or a CoW layer on top of the rootfs), move the data across nodes, etc.

Kosong: Kimi AI's Agent SDK

diptanu — Sun, 09 Nov 2025 05:44:42 +0000

Article URL: https://github.com/MoonshotAI/kosong

Comments URL: https://news.ycombinator.com/item?id=45863229

Points: 1

# Comments: 0

New comment by diptanu in "Benchmarking the Most Reliable Document Parsing API"

diptanu — Thu, 06 Nov 2025 21:24:18 +0000

There was an unusual traffic spike around that time, if you try now it should be a lot faster. We were calling up but there was not enough GPU capacity at that time.

New comment by diptanu in "Benchmarking the Most Reliable Document Parsing API"

diptanu — Thu, 06 Nov 2025 19:53:11 +0000

We haven’t tested Chandra yet, because it’s very new. Under the hood Tensorlake is very similar to Marker - it’s a pipeline based OCR API, we do layout detection, Text Recognition and Detection, Table Structure Understanding, etc. We then use VLMs to enrich the results. Our models are much bigger than marker, and thus takes a little longer to parse documents. We optimized for accuracy. We will have a faster API soon.

New comment by diptanu in "Benchmarking the Most Reliable Document Parsing API"

diptanu — Thu, 06 Nov 2025 19:50:02 +0000

It does, we have users in Europe and Asia using it with non English languages. Can you please send me a message at diptanu at tensorlake dot ai, would love to see why it didn’t work.

New comment by diptanu in "Benchmarking the Most Reliable Document Parsing API"

diptanu — Thu, 06 Nov 2025 19:48:55 +0000

OP mentioned Gemini and not Google’s Vertex OCR API which has very different performance and accuracy characteristics than Gemini

New comment by diptanu in "Benchmarking the Most Reliable Document Parsing API"

diptanu — Thu, 06 Nov 2025 19:20:47 +0000

Hey! I am the founder of Tensorlake. We benchmarked the models that our customers consider using in enterprises or regulated industries where there is a big need for processing documents for various automation. Benchmarking takes a lot of time so we focussed on the ones that we get asked about.

On Gemini and other VLMs - we excluded these models because they don't do visual grounding - aka they don't provide page layouts, bounding boxes of elements on the pages. This is a table stakes feature for use-cases customers are building with Tensorlake. It wouldn't be possible to build citations without bounding boxes.

On pricing - we are probably the only company offer a pure on-demand pricing without any tiers. With Tensorlake, you can get back markdown from every page, summaries of figures, tables and charts, structured data, page classification, etc - in ONE api call. This means we are running a bunch of different models under the hood. If you add up the token count, and complexity of infrastructure to build a complex pipeline around Gemini, and other OCR/Layout detection model I bet the price you would end up with won't be any cheaper than what we provide :) Plus doing this at scale is very very complex - it requires building a lot of sophisticated infrastructure - another source of cost behind modern Document Ingestion services.

Roles and Intelligence for Individual Contributors

diptanu — Sun, 12 Oct 2025 16:42:43 +0000

Article URL: https://raees.me/blog/role-and-intelligence/

Comments URL: https://news.ycombinator.com/item?id=45559572

Points: 1

# Comments: 0

RAG isn't dead, the bar has gone up

diptanu — Tue, 19 Aug 2025 20:48:09 +0000

Article URL: https://www.tensorlake.ai/blog/advanced-rag

Comments URL: https://news.ycombinator.com/item?id=44956119

Points: 2

# Comments: 0

New comment by diptanu in "So you want to parse a PDF?"

diptanu — Mon, 04 Aug 2025 04:23:46 +0000

We parse PDFs to convert them to text in a linearized fashion. The use case for this would be to use the content for downstream use cases - search engine, structured extraction, etc.

New comment by diptanu in "So you want to parse a PDF?"

diptanu — Mon, 04 Aug 2025 04:19:56 +0000

Yeah we don't handle this yet.

New comment by diptanu in "So you want to parse a PDF?"

diptanu — Mon, 04 Aug 2025 04:19:35 +0000

Yes this! We training it on a ton of diverse document images to learn reading order and layouts of documents :)

New comment by diptanu in "So you want to parse a PDF?"

diptanu — Mon, 04 Aug 2025 04:18:44 +0000

There are many cases images are exported as PDFs. Think invoices or financial statements that people send to financial services companies. Using layout understanding and OCR based techniques leads to way better results than writing a parser which relies on the files metadata.

The other thing is segmenting a document and linearizing it so that an LLM can understand the content better. Layout understanding helps with figuring out the natural reading order of various blocks of the page.

New comment by diptanu in "So you want to parse a PDF?"

diptanu — Mon, 04 Aug 2025 00:13:35 +0000

Disclaimer - Founder of Tensorlake, we built a Document Parsing API for developers.

This is exactly the reason why Computer Vision approaches for parsing PDFs works so well in the real world. Relying on metadata in files just doesn't scale across different source of PDFs.

We convert PDFs to images, run a layout understanding model on them first, and then apply specialized models like text recognition and table recognition models on them, stitch them back together to get acceptable results for domains where accuracy is table stakes.

Show HN: Tensorlake-Ingest, Parse, and Orchestrate Documents for AI Workflows

diptanu — Thu, 15 May 2025 17:57:21 +0000

Article URL: https://www.tensorlake.ai/blog/announcing-tensorlake-cloud

Comments URL: https://news.ycombinator.com/item?id=43997563

Points: 4

# Comments: 1