pip install shiftgate
Joined October 2022
- Tweets 300
- Following 832
- Followers 119
- Likes 859
26 Photos and videos
Pinned Tweet
19 Apr 2025
Got a cool opportunity to speak at the @supabase meetup on a hands-on setup and it's integration with @streamlit apps. Big thanks to @PrashantAbbi @baskaran1073 for the opportunity!
Special shout out to @SiddhantSadangi for the awesome streamlit-supabase connector.
19 Apr 2025
What a @supabase meet-up! 🚀
Co-organising this is wonderful, we had 50 attendees, tech sessions, and hands-on demos—like connecting Supabase MCP with @windsurf_ai.
Showcases included a dating app & a second brain using vector DB.
Supabase cheat sheets
Just the beginning for Supabase India! 🇮🇳
@CloseFutureHQ @ZeroicHQ
1
4
18
2,274
Had a great meetup at Mango Maxxing organized by @niketrajdwivedi ..
A great meetup with many researchers, builders, and their products. Got a chance to pitch and get feedback..
3 members are in@distilbok to try and include it in their workflow..
Let's see how it goes.
Special thanks to @niketrajdwivedi and @Ozacle23
Had special meetings with @itisRazak , @AbhiYanampally , @nkhanna94 , @suparnojit26
,@uncommonnayana , and @hah_serenity
These dudes are so high in energy that we can't miss them at all.🩷
5
4
29
1,380
I just cancelled my movie ticket to obsession because i vibecoded my version of it with fable for $250,000
1
3
164
razak wants to go to sf retweeted
Jun 13
We just wrapped up our @CloudflareDev buildathon and it was a full-house!
Thank you for showing up and shipping with us! We should do this more often, shouldn't we? 👀
Shoutout to @cengines_studio for letting us use their beautiful office
2
4
56
3,884
Coolest launch video in recent times, @googrish imma bombard castform with my curated cot data rn lesgoo
“don’t train your own model” is common ai advice. it's wrong. your token bill's the proof.
today, we’re excited to launch castform into open preview. castform is the easiest way for you to train your own model, on your own data.
open-weights models are performant and much cheaper. when trained on your task & proprietary data, they beat closed models. the thing standing between you and that was weeks of plumbing & years of ml expertise.
with castform, model training is as simple as prompt engineering. @castformai
bring your agent traces or raw corpora. castform turns it into training data, picks the right algorithmic recipes, manages gpus, and gives you an ide to watch and chat with your model as it learns.
see what you can build with castform👇
1
3
404
razak wants to go to sf retweeted
Jun 9
i am a founder looking for funding.
whom should I pitch? 👀
65
2
114
14,056
razak wants to go to sf retweeted
Jun 8
Hell yeahhh! Had a lot of fun building here ✨
Still looking for a full-time role btw. I ship full stack: agents, web, mobile. Fully agentic workflows: I steer, agents execute.
If you're hiring, please DM me. Would love to chat!
Jun 7
giving away $40k worth of prizes to the winners for ://agent_arena hackathon hosted at @localhostIND
1. @shashwatvalid
2. @pingrishabh
3. @plutodunk
4. @sauhard_07
5. @harishfelloff
special shoutout to our runner ups @NabhagMotivaras @heysagnik
6
11
1,498
Last week @cerebras put Multi-LoRA in private preview, leaving routing to app developers.
shiftgate is that layer: semantic routing to the right adapter per query.
When Multi-LoRA goes public, it just works. Interface identical.
pip install shiftgate — 3k downloads in a week
Multi-LoRA is in private preview on Cerebras Inference.
Deploy one base model alongside a library of LoRA adapters. Switch between them per request, with no reloading, no separate deployments, and no latency cost.
Available now for dedicated endpoint users. Reach out to your account rep to get access.
3
1
6
381
86
razak wants to go to sf retweeted
Last week @cerebras put Multi-LoRA in private preview, leaving routing to app developers.
shiftgate is that layer: semantic routing to the right adapter per query.
When Multi-LoRA goes public, it just works. Interface identical.
pip install shiftgate — 3k downloads in a week
Multi-LoRA is in private preview on Cerebras Inference.
Deploy one base model alongside a library of LoRA adapters. Switch between them per request, with no reloading, no separate deployments, and no latency cost.
Available now for dedicated endpoint users. Reach out to your account rep to get access.
3
1
6
381
From being first hand operator in factories to managing thousands at front , things always had a learning curve when it came to data , now as @zelantrixx is transitioning into human embodied intelligence, couldn't be more excited to see what frontiers can we tap into :)
1
2
132
razak wants to go to sf retweeted
Jun 3
🚀 Gemma 4 12B is here!
We partnered with @GoogleDeepMind to bring and optimize their new dense and unifed multimodal model for Apple Silicon.
◈ 12B dense · 256K context
◈ Thinking mode (built-in reasoning)
◈ Vision: dynamic res, OCR, UI charts
◈ Native audio: ASR speech translation
◈ Function calling for agents
◈ Text image audio, interleaved
Runs local. Get started now ⚡
> uv pip install -U mlx-vlm
github.com/Blaizzy/mlx-vlm
Jun 3
Meet Gemma 4 12B!
A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.
Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇
52
143
1,408
178,434
razak wants to go to sf retweeted
Jun 1
I'm moving on from my role as Head of Growth at @socapinc and looking for my next thing!
Built the distribution network for some of the best startups of this generation, did product launches that broke the internet, turned eyeballs into revenue.
I also build anything that catches my whimsy and have an unreasonable hatred for doing anything manually more than twice.
Did the 1→10 for others, now ready for the 0→1. Growth/product/founder's office at something early stage where I can own outcomes end-to-end.
Open to Bangalore or Remote. DMs open🤝
24
14
113
28,050
wow, this looks crazy, agents having their own phone!!
shoutout to @vlad_pliner and team for this producing this banger
Jun 1
There’s only one thing AI agents can’t do: reach the real world.
Should it stay this way?
No. That’s why we built Saperly.com
The first phone carrier for AI agents, enabling every agent to act in the real world.
1
3
220
Demoed shiftgate at ai weekends.
Started as a weekend experiment after reading the LORAUTER paper.
Now it can:
- Route prompts to the right LoRA
- Work with vLLM/Ollama
- Verify loaded adapters
- Run health checks with shiftgate doctor
Repo: github.com/mdrazak2001/shift…
2
1
15
1,026
razak wants to go to sf retweeted
May 31
ai & weekends ep 14, bangalore 🧚♀️
6
4
54
6,111