ninjaHattori1 (@Sahil100X) header
ninjaHattori1 (@Sahil100X) profile picture

Super 30 3.0 @100xSchool Remote AI Engineer @ Sift AI CF 1841

india
Joined January 2026
Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)Photo by ninjaHattori1 (@Sahil100X)
ninjaHattori1 retweeted
hi, i'm deepanshu. i build AI products end to end: voice agents, coding agents, and an exchange engine from scratch. looking for my next full-time role. under a minute of proof RTs mean a lot #HireDeepanshu
10
11
23
899
ninjaHattori1 retweeted
Git leaves CHERRY_PICK_HEAD on disk while the post-commit hook runs, so Graphify was skipping the rebuild on every clean cherry-pick. I fixed that a few weeks back, and XERJ just added the fix to their Corpus Hub for coding agents to reference.
Graphify just passed gstack on all time GitHub stars. Under 7 months since open source launch. Love to see this @garrytan
9
10
14
510
ninjaHattori1 retweeted
I built my own @emergentlabs from scratch. The AI agent was the easy part. The hard parts: - two servers fighting over one sandbox - a @daytonaio sandbox dying mid-build - agent events hammering Postgres How I solved each one, in 4 min 👇 How would you handle the crash?
Creating a distributed Lovable. Finalized the design and already started working on it. Tasks completed: - Wrote the streaming agent loop. - Added the question tool and other CRUD tools. - Added the sandbox startup code.
14
9
36
1,080
ninjaHattori1 retweeted
S30 folks keep shipping loved the detailed breakdown of the project . @waiting4_asi should check this out for next EP Cracked guy if you looking to hire!
This clinic receptionist isn’t human. I built her. Here’s how 👇 → LiveKit + WebRTC → AssemblyAI → DeepSeek → Rime → State machine orchestration → Appointments, insurance & refills → Warm transfers to staff → Provider failover → Postgres + Redis
2
11
371
ninjaHattori1 retweeted
This clinic receptionist isn’t human. I built her. Here’s how 👇 → LiveKit + WebRTC → AssemblyAI → DeepSeek → Rime → State machine orchestration → Appointments, insurance & refills → Warm transfers to staff → Provider failover → Postgres + Redis
17
10
3
62
4,674
ninjaHattori1 retweeted
I am Rohan building from more than 2+ years in @solana Built in more than 8+ Solana hackathons winning tracks in 2 of them . Also Contributing to open source projects like Rust Compiler , Agave etc. Proof of work - rohanbuilds.xyz/
I'm sponsoring 4 mini grants through @slate_fund for students and devs who want to be at solana summit mumbai on nov 2 but can’t afford the trip. what you get ~$500 each (flights and accommodation), paid directly in USDC who can apply - students or developers - actively building on @solana? you’ve got the best shot how to apply quote this post with a 15 to 30 second video covering: - who you are and where you’re from - what you’re building (link, repo, anything) - why the summit matters for you right now deadline 27 oct. four picked and posted the same day. know someone who should apply? tag them.
11
10
3
66
3,526
ninjaHattori1 retweeted
I built a dating app where AI agents go on dates for you. Two AI personas chat. A third AI judges compatibility. Hardest part? Making the conversations resumable without duplicate turns. Would you trust AI with your first date? 👀
10
8
13
733
ninjaHattori1 retweeted
FOMO to a builder
16
12
1
49
1,326
ninjaHattori1 retweeted
forgot to update here: Rescho just crossed 20+ users 😭 small number, but damn seeing real people actually use something i built feels different. rescho.deepanshu.live/
you know that 20 min argument of “bhai kaha khana hai?” 😭 so i built Rescho, basically Tinder but for restaurants. you and your friend join the same room, swipe nearby restaurants, and when both like the same one, boom match. built the whole thing realtime with Socket.io + Foursquare. swipe cards, room codes, match screen, all of it. try it: rescho.deepanshu.live
6
5
12
1,040
ninjaHattori1 retweeted
LUNAR NODES designs, builds, deploys and maintains software for founders, businesses and enterprises. Featured: Orin → orin.ashutoshsao.com/ We build it, we ship it, and we keep it running. lunarnodes.co.in/ DM us to build a production-ready app that just works. @ashutosh_sao @Deepjyoti100x @Adarsh_VMore @AnkurJangid02 @Sahil100X @deepanshuliv @jitendar100x @apaarmeet @byanujpatel @the_niresh
9
10
3
18
1,517
ninjaHattori1 retweeted
you know that 20 min argument of “bhai kaha khana hai?” 😭 so i built Rescho, basically Tinder but for restaurants. you and your friend join the same room, swipe nearby restaurants, and when both like the same one, boom match. built the whole thing realtime with Socket.io + Foursquare. swipe cards, room codes, match screen, all of it. try it: rescho.deepanshu.live
7
6
1
13
1,613
ninjaHattori1 retweeted
if your aura is high enough you don’t actually need a resume, they just hire you
127
1,520
252
15,406
408,042
ninjaHattori1 retweeted
Replying to @Sahil100X
@Sahil100X bro is live with the OG @kirat_tw , hopin to learn about Inference Engineering
8
5
3
65
5,105
Voice agents are in high demand. Here he explains them simply.
Built a Voice AI Agent for Plantiff Firms intake calls, completely from scratch. Here’s a breakdown of the entire architecture: → Cascading vs. Speech-to-Speech vs. Hybrid models → WebRTC with LiveKit for real-time voice → Silero VAD + Deepgram STT/TTS → Agent orchestrator built around a state machine → Intent detection + agent handoffs → Triage → Qualification → Appointment Booking → Conversation State vs. Task State
2
10
172
12,988
ninjaHattori1 retweeted
I cut Claude Code's context tokens by 87%. The benchmark I'm proudest of is the one where my tool loses by 293%.
34
36
5
392
45,988
ninjaHattori1 retweeted
Keep up the consistency bro @Sahil100X
I wrote GPU kernels for GPT2 124M inference engine from scratch in ~1500 lines of CUDA. Here I explain it simply.
1
1
6
517
ninjaHattori1 retweeted
Godd damm !!
I wrote GPU kernels for GPT2 124M inference engine from scratch in ~1500 lines of CUDA. Here I explain it simply.
2
1
45
2,655
ninjaHattori1 retweeted
Will be writing binary after sometime💀
I wrote GPU kernels for GPT2 124M inference engine from scratch in ~1500 lines of CUDA. Here I explain it simply.
1
1
8
258
ninjaHattori1 retweeted
From running inference at CPU to running it at GPU. Bro is stacking skill like side quests 🤯
I wrote GPU kernels for GPT2 124M inference engine from scratch in ~1500 lines of CUDA. Here I explain it simply.
1
8
238
20,141
ninjaHattori1 retweeted
Bro is building crazy stuff, day by day
I wrote GPU kernels for GPT2 124M inference engine from scratch in ~1500 lines of CUDA. Here I explain it simply.
3
26
615
81,297