Nishchal Marur

Machine Learning Engineer II
Computer Vision & 3D
MSML at UMD

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[AI at Meta @AIatMeta]

We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2.

Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication.

We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating.

🧵👇

[S.🎧 @1ssve]

Job applications really be like:
“Upload your resume… now type your resume… now explain your resume.”

[Jahir Sheikh @jahirsheikh8]

If it’s real-time → WebSockets
If it’s scale → Kafka
If it’s simplicity → REST
If it’s chaos → GraphQL
If it’s AI → Python
If it’s infra → Go
If it’s logs → ElasticSearch
If it’s low-latency → Redis
If it’s high-availability → Postgres
If it’s streaming → Flink
If it’s low-level → C
If it’s high-performance → C++
If it’s enterprise → Java

If it’s frontend → React
If it’s styling → Tailwind
If it’s fullstack → Next.js
If it’s backend → Node.js
If it’s type safety → TypeScript
If it’s auth → OAuth
If it’s payments → Stripe
If it’s search → Meilisearch
If it’s caching → CDN
If it’s queues → RabbitMQ
If it’s containers → Docker
If it’s orchestration → Kubernetes

If it’s monitoring → Prometheus
If it’s dashboards → Grafana
If it’s CI/CD → GitHub Actions
If it’s version control → Git
If it’s testing → Jest
If it’s API testing → Postman
If it’s secrets → Vault
If it’s messaging → gRPC
If it’s event-driven → Pub/Sub
If it’s data warehouse → BigQuery
If it’s vector DB → Pinecone
If it’s serverless → AWS Lambda

[Akshay 🚀 @akshay_pachaar]

8 AI model architectures, visually explained!

Everyone talks about LLMs, but there's a whole family of specialized models doing incredible things.

Here are their architectures mapped out:

[Nishchal Marur @Nishchalmn]

@huang_chao4969 Great work! Although I am curious how different or better it is than Google's @NotebookLM

[Akshay 🚀 @akshay_pachaar]

8 RAG architectures for AI Engineers:

(explained with usage)

1) Naive RAG
↳ Retrieves documents purely based on vector similarity between the query embedding and stored embeddings.
↳ Works best for simple, fact-based queries where direct semantic matching suffices.

2) Multimodal RAG
↳ Handles multiple data types (text, images, audio, etc.) by embedding and retrieving across modalities.
↳ Ideal for cross-modal retrieval tasks like answering a text query with both text and image context.

3) HyDE (Hypothetical Document Embeddings)
↳ Queries are not semantically similar to documents.
↳ This technique generates a hypothetical answer document from the query before retrieval.
↳ Uses this generated document’s embedding to find more relevant real documents.

4) Corrective RAG
↳ Validates retrieved results by comparing them against trusted sources (e.g., web search).
↳ Ensures up-to-date and accurate information, filtering or correcting retrieved content before passing to the LLM.

5) Graph RAG
↳ Converts retrieved content into a knowledge graph to capture relationships and entities.
↳ Enhances reasoning by providing structured context alongside raw text to the LLM.

6) Hybrid RAG
↳ Combines dense vector retrieval with graph-based retrieval in a single pipeline.
↳ Useful when the task requires both unstructured text and structured relational data for richer answers.

7) Adaptive RAG
↳ Dynamically decides if a query requires a simple direct retrieval or a multi-step reasoning chain.
↳ Breaks complex queries into smaller sub-queries for better coverage and accuracy.

8) Agentic RAG
↳ Uses AI agents with planning, reasoning (ReAct, CoT), and memory to orchestrate retrieval from multiple sources.
↳ Best suited for complex workflows that require tool use, external APIs, or combining multiple RAG techniques.

👉 Over to you: Which RAG architecture do you use the most?

[Mayank Vora @aiwithmayank]

OPENAI, ANTHROPIC, AND GOOGLE DON’T PROMPT LIKE YOU

They use internal techniques that turn LLMs into precision machines
Accuracy jumps. Hallucinations drop.

Here are 10 of those techniques (Bookmark this for later):

[Nishchal Marur @Nishchalmn]

@piyascode9 Agent

[Louis Gleeson @aigleeson]

🚨 BREAKING: HuggingFace just released 9 free AI courses.

You can now learn LLMs, vision, agents & more all for free.

Here’s how to start today ↓

[Nishchal Marur @Nishchalmn]

@krithik255 @Flipkart Problems like these happen more in Flipkart compared to Amazon.

[Nishchal Marur @Nishchalmn]

@beebomco Google Pixel 4a because it has nailed the basics right!

[Nishchal Marur @Nishchalmn]

@krithik255 @MKBHD MKBHD always hyping around apple products😑...

[Nishchal Marur @Nishchalmn]

@krithik255 Gottilvala🤔

[Beebom @beebomco]

A new camera has been developed in China that uses a lidar-based system and can click a picture of you from 45 kilometers away.

(Not developed by Huawei.)

[taran adarsh @taran_adarsh]

#AvengersEndgame emerges a game changer... Sets new benchmarks on Day 1... An eye opener for the industry that feels national holidays/festivals yield best results at the BO... Fri ₹ 53.10 cr Nett BOC. India biz. Gross BOC: ₹ 63.21 cr. NON-HOLIDAY RELEASE.

[Nishchal Marur @Nishchalmn]

@TechnicalGuruji Poco F2?

[Nishchal Marur @Nishchalmn]

@krithik255 @LGIndia 😂😂😂