Foundational Data Layer for AI: Combine scalable vector search with memory-optimized graph and multimodal data management
Joined November 2018
- Tweets 590
- Following 29
- Followers 404
- Likes 17,379
173 Photos and videos
ApertureData retweeted
I built an enterprise memory engine with an AI that kept forgetting things.
Meet Claudette: brilliant coder, relentless collaborator, and quietly allergic to integration tests.
The raw reality of building Aperture Nexus:
• The Mock Trap: Claudette built tests that only tested themselves.
• The Unfair Advantage: Why a unified multimodal backend (@ApertureData ) saved us months of infrastructure engineering.
• Context Drift: Needing the very tool I was building just to resolve cross-LLM entity data and keep my data states straight.
If your AI tools feel "brilliant but amnesic", this "behind the scenes" engineering story is for you.
👇 Read the full breakdown here:
medium.com/@vishakha041/me-m…
2
6
1,035
ApertureData retweeted
Thanks @Coffee_and_NLP
We use ApertureDB from @ApertureData
It already supports vector search , graphs and is multimodal native. Removes a lot of infra complexity from the cognition layer. Memory is open source. DB has community edition and then enterprise scaling
1
2
42
ApertureData retweeted
Multimodal-native data and memory infrastructure for enterprise AI
@ApertureData
cloud.aperturedata.io/signup
1
1
55
ApertureData retweeted
3
5
815
ApertureData retweeted
Kicked off @Garageplusepoch accelerator today in Taipei with 14 other teams selected We are happy to represent @ApertureData at the meetings and @computextaipei The energy around AI data management and agent memory infrastructure is unmistakable and we are going to discuss how to scale this in production
1
3
6
295
May 6
Something we understand very well here!
The Integration Tax: Relying on pluggable, fragmented backends is a performance trap.
We break down why "pluggable" often means "inconsistent" and why the win for 2026 is moving the complexity from the application layer to the data layer. (4/5)
2
56
On episode 54 of Generationship, @rachelchalmers is joined by Vishakha Gupta (@vishakha041) of @ApertureData to discuss how AI systems break down in the real world, and how to fix them. They dive into multimodal data, graph databases, and the challenges of unifying complex data pipelines for enterprise use. Tune in!
hubs.ly/Q04bQ0Ch0
2
2
96
ApertureData retweeted
The real hurdle going forward? Moving from Storage to Memory.
Agents don't just need a place to store & retrieve data; they need a unified engine for Vector Graph Multimodal context and memory in real-time.
That’s the "cognition infrastructure" we are building at @ApertureData (3/4)
1
1
2
74
On episode 54 of Generationship, @rachelchalmers sits down with Vishakha Gupta (@vishakha041) of @ApertureData to examine the intersection of AI, data infrastructure, and human memory. They explore how multimodal data, graph structures, and vector search must come together to support next-generation AI systems. Tune in!
hubs.ly/Q04bL4Y30
2
3
128
ApertureData retweeted
Understanding "What" is out there is step one. Step two is the "How."
In Part 2c, we will share our observations and guide on how to select, and ask the 8 critical questions every architect must ask before committing to a Cognition stack.
Full Audit here: aperturedata.io/resources/ai…
1
2
30
ApertureData retweeted
I spent the last few weeks auditing 20 AI memory frameworks.
The good news: The "Digital Attic" era is ending. We’re finally seeing real work on connective tissue and reasoning layers.
The bad news: We’re trading one bottleneck for three new ones. 🧵
1
1
2
69
Apr 24
The discussion around surfacing dark data and the improvements in multimodal models is where ApertureDB can fit in and fill the gap with it's support for different modalities, efficient searches, and the ability to build a scalable agentic memory layer on top!
At @googlecloud Next '26, Google has gone all in on Gemini Enterprise, launching new chips, agentic cloud, more advanced multimodal models, agentic security, testing and SDLC for this new world of agents, making most of dark data, and so much more. There were also some very different and interesting things that caught my eye walking around...
All those attending GCP, what are your observations from this year?
#GoogleNext
2
56
On episode 54 of Generationship, @rachelchalmers sits down with Vishakha Gupta (@vishakha041) of @ApertureData to explore the hidden infrastructure challenges behind modern AI. They unpack why multimodal data systems are still fragmented, how graph and vector approaches can be unified, and what it takes to build production-ready AI pipelines.
hubs.ly/Q04bGzs_0
2
3
280
ApertureData retweeted
Had a great time chatting with @rachelchalmers about multimodal systems, memory, and the gaps we’re still working to close in AI infrastructure.
@ApertureData
On episode 54 of Generationship, @rachelchalmers sits down with Vishakha Gupta (@vishakha041) of @ApertureData to explore the hidden infrastructure challenges behind modern AI. They unpack why multimodal data systems are still fragmented, how graph and vector approaches can be unified, and what it takes to build production-ready AI pipelines.
hubs.ly/Q04bGzs_0
1
1
2
114
ApertureData retweeted
Which AI memory framework should I pick? Well… glad you asked 👇
Here is the Part 2a of our landscape study with @alinahm from @tribecap : The Path to Machine Cognition.
We’re using the KMC Blueprint to help you make sense of the 2026 market:
- Knowledge (the library)
- Memory (the growth layer)
- Context (the connective tissue)
Stop just 'storing data' and start building Cognition.
Stay tuned for Part 2b coming next week: The full analysis of 20 frameworks, from local bots to the Memory Mesh!
aperturedata.io/resources/th…
2
3
165
Apr 9
Is context just a decision trace, or is it something more?
At @GoogleCloud Next, we’re moving past the "vector search" hype to talk about the actual architecture of AI Cognition. If you're building agents intended for production, the "digital attic" approach to memory won't scale. #GoogleCloudNext
1
1
3
44
Apr 9
Knowledge isn't just text. It’s the relationship between a technical schematic, a transcript, and a product specification.
Fragmenting vectors, metadata, and data into different silos creates a massive "data tax" in latency and complexity. The future of the stack is unified, multimodal, and graph-aware.
1
2
17
Apr 9
The next generation of reasoning agents needs a persistent, evolving knowledge, memory, and concept structure, not just a "fetch" command.
We’re at #GoogleCloudNext to discuss building this cognitive layer for the enterprise. Let’s find a whiteboard and get into the systems architecture of it.
DM @vishakha041 to find a corner to chat.
1
34
Apr 8
Adding ApertureDB as a substrate to building this!
I’m currently finalizing a deep dive into ~20 AI memory frameworks and research projects.
But before the data drops, we need to talk about why most "memory" frameworks aren't actually building memory.
They're just building better search engines. 🧵
2
28