Amisha Kirti

[Amisha Kirti @kirtiastic]

Thankyou for the mention! @GoogleDevsIN Looking forward to more contributions! 💙

[Google for Developers India @GoogleDevsIN]

Our DevLibrary contributors have outdone themselves at the Google #DeveloperStudentClubs #FlutterFestivals. 👏🏻

Do you want to #DevelopWithGoogle?
If yes, get your hands on the submissions of the Flutter Libraries usage and get started with your own #Flutter journey.

[Karan🧋 @kmeanskaran]

Stop building AI/ML projects in 2026 like it's 2023

Do this instead.

1. Classic ML:
- Stock forecasting:
Use Yahoo Finance API for stock forecasting. Learn moving averages, seasonality, and other features. Apply statistics, feature engineering, and model training.

Use transfer learning (index as parent model, stock as child model). Study ML techniques in finance and implement them.

Perform model versioning, experimentation, and MLOps lifecycle with Docker, CI/CD, and drift detection.

- Image/Video classification:
Pick static data from Kaggle. Perform data annotation, augmentation, and class imbalance handling. Train neural networks and explore various techniques.

Understand quantization and inference on NVIDIA Triton Inference. Deploy a lightweight model on mobile devices with better latency.

Build a pipeline to switch to improved model versions and monitor performance plus GPU usage.

---

2. Complete VLM/LLM pipeline:
- Design API for LLM:
Take any model from Unsloth. Understand fine-tuning stages, data cleaning, and chat templates.

Run fine-tuning on multi-GPU setup using PEFT. Save quantized versions on Hugging Face. Load the model in vLLM inference and build a complete FastAPI backend.

Fine-tune two models (one for text generation, one for reasoning). Route user queries with KV-cache and rate limiting.

Deploy the API with authentication on AWS/GCP. Monitor backend, tokens, and GPU usage. Stress-test the architecture and fix issues.
Excellent project.

---

3. Multi-Agent System
- Agent Harness with Ops (currently building):
Design a multi-subagent system that operates in isolation. Add a memory layer for conversations.

Learn prompt caching vs semantic caching tradeoffs. Build a backend with Celery workers to handle requests without exhausting the LLM. Evaluate agent orchestration with prompt versioning.

Deploy the orchestration backend with Docker, observability, and Kubernetes.

Keep in mind for any project:
- Learn architecture design and flow.
- Use rate limiting, caching, and Docker every time.
- For LLM projects, track token usage and design session limits.
- Deployment is mandatory on AWS or GCP free tier.
- For mid-to-senior roles, focus on Kubernetes, load balancing, multi-GPU training, and inference layers over data and project titles.

Instead of wasting $100 on certificates, spend on cloud credits and API keys for hands-on end-to-end design.

Lastly, project design and system understanding matter more than fancy titles or LLM providers.

Always ask Claude and Chatgpt for peer coding and designing. Simulate QnA sessions for deeper understanding.

Keep learning ;)

[Param @ParamSiddh]

1) I switched to AI Engineering 1 years ago!
It was the best career move I ever made.

If you want to start today, here's a roadmap:

[Ashwin Nair 🎴 @AshwinN14729359]

@elonmusk Geoffrey Hinton had a vision before the world could even imagine it.

[Eduardo Ordax @ordax]

Great resource on Agentic AI by Weaviate. I recommend to to download the tutorial.

Link: https://t.co/iWecDlkcDK

[Eduardo Ordax @ordax]

I saw a guy coding today:

Tab 1 ChatGPT.
Tab 2 Gemini.
Tab 3 Claude.
Tab 4 Grok.
Tab 5 DeepSeek.

He asked every AI the same exact question.

Patiently waited, then pasted each response into 5 different Python files.

Hit run on all five.
Pick the best one.
Like a psychopath.

It's me.

[Eduardo Ordax @ordax]

My wife asked me why I speak so softly in the house.

I said I was afraid Mark Zuckerberg was listening.

She laughed.
I laughed.
Alexa laughed.
Siri laughed.
My phone laughed.
The Tesla next door laughed.

We are so cooked.

[Karan🧋 @kmeanskaran]

best LLM meme so far😂😂
@ordax

[Sam Altman @sama]

It is a very smart model, and we have come a long way since GPT-5.1:

[Amisha Kirti @kirtiastic]

AI Agents vs Agentic AI

https://t.co/Ww4Z8Iblgw

[Rishikesh📱 🏊‍♂️ @RustyRishii]

Found a website that allows you to read paid medium articles for free

https://t.co/bSSGc2LQ85

[NK @Thereal_Nsikak]

“A murderer is not the first person to see a victim dead but the last one to see them alive.”

[Viक़as @VlKAS_PR0NAM0]

Goose Bumps 🔥

Hats off to this gentlemen, Now Pakistani must know how Ruthless is Modiji🇮🇳

[Amisha Kirti @kirtiastic]

Tuesdays are the real Monday!! 💁

[World updates @itswpceo]

Mind Blowing International Borders From Around the World

A Thread 🧵

1. Netherland & Belgium

[Gabbar @GabbbarSingh]

Flow of events:

- Pahalgam attack. 26 civilians brutally killed.

- India attack 9 terror bases. 24 missiles. None intercepted.

- Pak attempted to bomb 15 cities across North & West India. EVERY missile/drone intercepted

- India retaliate by bombing nearly every city of Pakistan by Kamikaze drones. Destroy their Air defence system. Lahore + 2 more.

[Ahmed Khaleel @ahmedkhaleel04]

visualize any codebase 👀

[Sam Altman @sama]

it's super fun seeing people love images in chatgpt.

but our GPUs are melting.

we are going to temporarily introduce some rate limits while we work on making it more efficient. hopefully won't be long!

chatgpt free tier will get 3 generations per day soon.

[Amisha Kirti @kirtiastic]

@prince_h_hm Corporate se door 😂

[Amisha Kirti @kirtiastic]

So annoyed right now — in dire need of a private jet. 🥱

[Amisha Kirti @kirtiastic]

@gunsnrosesgirl3 Jugnu (Firefly)