Vaibhavi Gangwar
@vaibhavi0601Building @getmaximai, previously @google
Similar User
@suyash666
@Smitgupta
@surajku89969190
@aimar09
🧵 Red Teaming in AI Red teaming is a critical practice in AI safety. It involves testing LLMs to find vulnerabilities, such as generating content that violates norms, policies, and rules during their safety training. Red teaming typically involves experts manually probing…
Exciting news! @ManavSinghal157 from our team is presenting his work on NoFunEval, a benchmark evaluating code LMs on non-functional requirements, at @COLM_conf Connect with Manav to discuss evaluations and code LMs! #COLM24
Really excited to be presenting our work NoFunEval (benchmark evaluating code LMs on non-functional requirements) at @COLM_conf Drop by our Poster 4 on Tuesday 8th from 4:30pm. Hit me up if you want to catch up to discuss more about evaluations or code LMs!
@ManavSinghal157 and @curiousZeedX are cooking up something cool for @getmaximai We're tackling the complexity of agentic workflows head-on, focusing on streamlining build and evaluation processes. Our goal? Making continuous evaluation a breeze.
Phew! I just wrapped up our monthly recap, and wow - we've been busy (so busy that I skipped writing these for a few months)! 🚀. We've shipped a ton of cool stuff lately. 💪
🧵Understanding Human and LLM Preferences 1/ 🧠 In a study, researchers analyzed preferences from both humans and 32 different LLMs, using real-world user-model conversations. This fine-grained, scenario-wise analysis revealed some compelling insights. #LLMs #AI #ChatGPT
Evalutaion Nugget 🤔 1/6 📷 Exploring multi-vector rerankers like ColBERT: These models blend bi-encoder efficiency with cross-encoder depth. For instance, ColBERT precomputes document representations but enriches query-document interaction during similarity computation. #AI…
Since Anthropic released Claude 3.5, there has been a constant buzz on Twitter. We have constantly seen arguments - GPT 4o this, Claude 3.5 Sonnet that... We ran experiments on our internal benchmarks. The details of each experiment are in the blog. blog.getmaxim.ai/claude-3-5-son…
We have been trying out @AnthropicAI B Claude 3.5 Sonnet internally and seeing impressive results.🙌 As we share more findings, sending a virtual hug to all the AI engineers🫡
📢1/7 Today, we are thrilled to announce the general availability of the Maxim AI platform getmaxim.ai. Since starting Maxim last year, we have been moving at an aggressive pace to empower AI developers to ship their products with speed and confidence. We are…
Congratulations @vaibhavi0601 @akshay_deo, @ElevCap and the entire @getmaximai team!
We are super excited to partner with @getmaximai as @vaibhavi0601 and @akshay_deo announce the fundraise and general availability launch of their enterprise-grade evaluation and observability platform, setting new standards for AI application development. With the recent…
We are super excited to partner with @getmaximai as @vaibhavi0601 and @akshay_deo announce the fundraise and general availability launch of their enterprise-grade evaluation and observability platform, setting new standards for AI application development. With the recent…
🔈 Investment Memo With the rapid growth of generative AI, the need for robust testing frameworks has never been more critical. @vaibhavi0601 (VG) and @akshay_deo's deep expertise in AI and developer tools, gained from their tenure at Google and Postman, gives us immense…
United States Trends
- 1. Jon Jones 223 B posts
- 2. Jon Jones 223 B posts
- 3. #UFC309 331 B posts
- 4. Chandler 91,4 B posts
- 5. Aspinall 26,1 B posts
- 6. Good Sunday 51,8 B posts
- 7. #Jays_Neighborhood N/A
- 8. Oliveira 75,8 B posts
- 9. Kansas 24,1 B posts
- 10. Mike Johnson 48,2 B posts
- 11. #discorddown 7.358 posts
- 12. #ปิ่นภักดิ์ตอนจบ 1,35 Mn posts
- 13. Pereira 12,9 B posts
- 14. Alec Baldwin 9.535 posts
- 15. THE LOYAL PIN FINAL EP 1,17 Mn posts
- 16. Bo Nickal 9.406 posts
- 17. ARod 2.213 posts
- 18. Dana 273 B posts
- 19. Do Bronx 12,1 B posts
- 20. Mayu 18,4 B posts
Something went wrong.
Something went wrong.