Modern AI products need strong infrastructure for reliability, debugging, and latency management. Docsum leans on tools that make LLM workflows more legible at scale.

Langfuse — LLM observability

“One thing that we started using for kind of LLM observability is Langfuse.”

“Another YC S23 company. It’s been great.”

Langfuse helps the team inspect prompts, trace cost vs. quality tradeoffs, and run faster iteration cycles with less guesswork.

Hatchet — Asynchronous processing & orchestration

“Oftentimes, any LLM call is long-running, and so we use Hatchet for all of this type of orchestration.”

Hatchet enables multi-step processes without blocking core application performance — critical for long-running analysis workflows inside Docsum’s contract review pipeline.

AI isn’t just a feature in Docsum’s product, it’s a core operating primitive for how the team collaborates.

Claude (Enterprise + Claude Code)

“We use Claude Code, and then we use Enterprise Claude for everybody on the team.”

“We’ve been pretty diligent about getting every member of the team to be either working in a project or publishing a skill… such that any bit of work that we use can be transferable to somebody else.”

“Our perspective is that if you’re prompting something and it’s not in a project, it’s almost like not contributing to our ability to collaborate as a team.”

Engineers use Claude Code for exploratory work and refactoring. Non-technical teammates use Enterprise Claude to eliminate handoffs, spec work faster, and prototype concepts independently.

The company avoids tool sprawl and optimizes for high-bandwidth communication across product, engineering, and design.

Linear — Product & engineering hub

“We try to consolidate tools as much as possible… All product and engineering documentation now lives in Linear.”

Linear acts as the structured layer where specs, tickets, and discussions stay tightly coupled.

Notion — Company docs

“We still have Notion for, like, company docs and things like that.”

Notion remains the home for onboarding, long-form docs, and operational knowledge that doesn’t belong in a ticketing workflow.

Docsum’s design process favors speed and clarity over high-fidelity ceremony early on.

Whimsical — UI, workflows, wireframes

“One tool that I really like using for design and prototyping… is Whimsical.”

“It is just so good. It’s so good, and it embeds well within Linear and then Notion as well.”

“We wrote a really good component library for our front end, and then all of the workflow and UI wireframing is done in Whimsical directly.”

Whimsical enables the team to quickly explore user journeys, interface patterns, and complex workflows before moving to production design systems.

Founders now operate as their own media studios — product demos, feature drops, and educational material all benefit from strong tooling.

Veed — Video editing

“I use Veed… for editing.”

Making content can be difficult and time-consuming, so Veed helps Docsum create quality content when they need to move fast.

ScreenStudio — Screen recording & capture

“I use ScreenStudio for capturing videos and screenshots.”

“It’s a really nice way of having a good presentation for that.”

These tools support everything from investor demos to customer onboarding to marketing content, keeping the feedback loop tight without external production overhead.

Shaunak Turaga is the Co-founder and CEO of Docsum, an AI-powered contract repository and negotiation system used by legal and revenue teams to accelerate deal cycles and gain clarity on their commercial relationships. Docsum combines AI-assisted redlining, contract analysis, and post-signature reporting to give companies visibility into risks, pricing, renewals, and obligations. Shaunak founded Docsum in 2023 and has been an active member of NYC Founders Club since 2024.