sandeeprdy1729 commited on
Commit
98c16b6
Β·
verified Β·
1 Parent(s): 3479d02

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +52 -7
README.md CHANGED
@@ -1,10 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
- title: README
3
- emoji: πŸ¦€
4
- colorFrom: pink
5
- colorTo: pink
6
- sdk: static
7
- pinned: false
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
  ---
9
 
10
- Edit this `README.md` markdown file to author your organization card.
 
1
+ # TIMPS β€” AI Research & Engineering. Think. Right.
2
+
3
+ **AI research and engineering for human intelligence.**
4
+
5
+ AI will reshape civilization. TIMPS exists to make sure it reshapes it for the better β€” for every one.
6
+
7
+ πŸ”— Website: [timps-website.vercel.app](https://timps-website.vercel.app/)
8
+ πŸ“„ GitHub: [github.com/timpss](https://github.com/timpss)
9
+ πŸŽ₯ YouTube: [@Timps-090](https://www.youtube.com/@Timps-090)
10
+ 🐦 X / Twitter: [@TIMPs_ai](https://x.com/TIMPs_ai)
11
+ πŸ’Ό LinkedIn: [TIMPS](https://www.linkedin.com/company/timps/)
12
+
13
+ ---
14
+
15
+ ## Featured Research
16
+
17
+ ### TIMPS-Coder-7B β€” 98.8% HumanEval, 4-stage training pipeline
18
+
19
+ A 7B-parameter code model fine-tuned through SFT, GRPO, DPO, and Self-Play.
20
+
21
+ - **MBPP+**: 73.3% β€” the clean, uncontaminated benchmark
22
+ - HumanEval results disclosed with full transparency on training-data overlap
23
+ - License: Apache-2.0
24
+
25
+ [Read the paper β†’](https://timps-website.vercel.app/timps-coder-7b.html)
26
+
27
  ---
28
+
29
+ ## Models
30
+
31
+ | Model | Description |
32
+ |---|---|
33
+ | **TIMPS-Coder-7B** | 7B code model, 4-stage training pipeline (SFT β†’ GRPO β†’ DPO β†’ Self-Play) |
34
+ | **TIMPS-Coder-0.5B** | Lightweight fine-tune of Qwen2.5-Coder for fast, local code generation |
35
+ | **TIMPS-Vision** | *(coming soon)* |
36
+
37
+ ## Products
38
+
39
+ - **[TIMPS Memory](https://timps-website.vercel.app/timps-memory.html)** β€” try it now
40
+ - **[TIMPS Swarm](https://github.com/Sandeeprdy1729/timps-swarm)** β€” multi-agent LangGraph framework with MCP integration
41
+ - **TIMPS-Coder** β€” our code model family
42
+
43
+ ---
44
+
45
+ ## Our Mission
46
+
47
+ At TIMPS, we build AI to serve humanity's long-term well-being β€” without harming any creature.
48
+
49
+ - **Think. Right.** β€” our founding philosophy
50
+ - **TIMPS Safety Constitution** β€” what our agents will never do
51
+ - **TIMPS Economic Index** β€” AI's impact on Indian livelihoods
52
+
53
  ---
54
 
55
+ Β© 2026 TIMPS Β· MIT License