📢 I've tried #NovitaAI for querrying large datasets
novita.ai/models/model...
✅ ️Pros: support for OpenAI API is highgly convenient
❌️ Cons: The rate limit (20 Req/min) was a blocker for efficient batch processing; this is too impractical
🌟 In few lines: github.com/nicolay-r/bu...
📢 If you around of handling massive amount of data with LLM and prompt schema, then you might be interested to have a look on bulk-chain: proxy in between batched data iterators and LLM providers:
📦️ The latest version is on PyPI:
pypi.org/project/bulk...
#data #iterator #cot
What's new:
☑️ Simplified inference setup
The API is now closer to the OpenAI paradigm for toggling streaming.
☑️ 🛠️ Fixed issues when passing code contain {} blocks
☑️ 🛠️ Async streaming + batching now works properly
☑️ 🛠️ Logging of prompts could be disabled
📢 Excited to share a new version of 🌟 bulk-chain 🌟!
Bulk-chain is high-level wrapper over LLM providers for efficient quering LLMs hosted by third-party services.
It brings native batching via support of async clients.
🌟 github.com/nicolay-r/bu...
Screenshot of YouTube channel "Fireplace 10 hours" showing a single uploaded video titled "Fireplace 10 hours full HD" with 156 million views, uploaded 9 years ago. The channel has 108K subscribers and only one video. The thumbnail and profile picture both show a cozy fireplace with flames.
Imagine uploading 1 YouTube video 9 years ago and still generating 2M+ views a month.
No algorithm hacks. No posting schedule. No "content strategy."
Just a fireplace burning for 10 hours.
This is what digital real estate actually looks like.
Framework is available:
github.com/nicolay-r/bu...
🧵3/n
#bulkchain
🔑 bulk-chain solves the following problems:
✅ Effective handling CoT schema with big amount of prompts and parameters they are based on (batching policies)
✅ Effective and easy-to-apply for data-iterators (datasets handling)
🧵2/n
#bulkchain
🚨 Time-Effective LLM Querying in Information Retrieval Tasks
At Research Colloquium in Technische Universität Chemnitz we presented a framework for time-effective data handling with prompting schemas👇
🎬 youtube.com/watch?v=pa8j...
🧵1/n
📢 If you do studies in Natural Language Processing, this may be of interest.The International Conference ICNLP 🇨🇳 is comming in 2026. This time it is going to be the 8th edition, and for those who interested in submission, call is open until 30'th of december
www.icnlp.net
🌟 Powered by:
🧑💻 bulk-chain (inference framework): github.com/nicolay-r/bu...
↗️ nlp-thirdgate (providers): github.com/nicolay-r/nl...
🧵4/n
🔑 We use separate providers, in particular supporting to @replicate.com .
Why it matters among the other tutorials: this is easy to extend with the other models supported by a particular platform or change to the other providers
🧵3/n
With the second part we making step from supporting streaming of data chunks towards ➡️ support inference from 3-rd party providers 🤖 (@openrouter.bsky.social , @replicate.com , etc.).
🧵2/n
🚨 The second part of the tutorial series on mimimalistic LLMs integration in web environment is here! 🎉
🎬️ youtu.be/yNKYJzlKxh0
🧵1/n
🔑 This features the no-string framework for quierrying LLMs in various modes: sync, async and with optional support for output streaming.
📦️ In the latest 1.2.0 release, the updates on outlining API parameters for inference mode.
🧵2/n
📢 For those who interested in applying LLM for inferring iterators of data with CoT / prompts, this update might be relevant. Deligted to share the new release of the bulk-chain. This contributes to efficient AI querying for synthetic data gen.
🌟https://github.com/nicolay-r/bulk-chain
🧵1/n
🤖To relay from LLMs in Sentiment Analysis, here are the sub-tasks that could be done by other non-LLM systems:
1. Text Translation: github.com/nicolay-r/bu...
2. Named Entity Recognition and Masking: github.com/nicolay-r/bu...
🧵5/n
📦To Use LLMs out-of-the box effective:
1. Few-shot and Fine-tuning: nicolayr.com?tab=professi...
2. THoR Reasoning: nicolayr.com?tab=professi...
🧵4/n
🔗 To opt-out from LLM, below is the list of the key resources that were introduced:
1. Conventional Frameworks: nicolayr.com?tab=professi...
2. NN and CNN: nicolayr.com?tab=professi...
3. NN Framework with Att: nicolayr.com?tab=professi...
4. Towards Self-Attention: nicolayr.com?tab=professi...
🧵3/n
🔑 Unlike similar previous talks presented in seminars at
@WolfsonCollege and @UofGlasgow, this one brings frameworks at survey level of numerous subtasks in sentiment analysis. This unites benchmarking on numerous data sets. 📊
Slides: drive.google.com/file/d/1JaWD...
🧵2/n
🚨 Can we opt-out from LLMs or we should use them in efficient way for Sentiment Analysis?
Answering to that question with the survey of previously developed frameworks in Sentiment Analysis. This become a lecture for CSC8644
@UniofNewcastle
🎬 youtube.com/watch?v=dYEx...
🧵1/n
29k participants at Neurips 2025
In further posts here you may expect to see key takaways in emerge of AI / Data and Healthcare out of keynote talks.
🧵4/n
#Birmingham #nhs #10yearplan #ai #data #genAI #appplication #healthcare
🌟 Highlights of subjects we contribute to:
📜 CASA2025: Using Chain-of-Thouhts for Textual Radiological Report Evaluation -- nicolayr.com?data_type=pa...
📜 CLEF2025: Clinical Textural Report Summarization -- nicolayr.com?data_type=pa...
🧵 3/n
🔑 My provision contribution to the event is attendance on the basis of being a Research Fellow at Bournemouth University. In collaboration with the medical team, we applying and advancing Natural Language Processing in Marking Medical Images and Radiological Reports Evaluation 🤖
🧵2/n
📢 Already head to the Birmingham to visit Delivering the NHS 10-Year Plan with AI & Data. Looking forward for keynote talks on AI adoption for Health servises, AI Risks.
🙌 If you're around or sharing mutual interest, let's connect!
🧵1/n
📢This year finds me with the very second attempt in Run Bournemouth Supersonic 10K. Thanks to every supporter on that breezy way 🏖️ 💨...
🤔 Did I smashed it?
⏱️ www.myrunning.uk/run-bournemo...
#runbournemouth #bournemouthac #supersonic #10k #bournemouth #bournemouthbeach
⚕️ The PubMed OA subset byshares a metadata for 35M articles. Suddenly, the existing article parser was supported until 2024
📢 Excited to share project, which contirbutes:
☑️ Downloader for files
☑️ No-string iterator over articles
github.com/nicolay-r/pu...
#MedLibs
📢 Excited to share that our previosly presented studies on clnical report summarization 📝🩺🧑⚕️ at #CLEF2025 has become published an available at CEUR-WS:
📄 Check out: ceur-ws.org/Vol-4038/pap...
📢 Just registered at #Manchester #Health Series and ready connect and dive into talks about healthcare 🩺 and AI.
Personal interest and scope: present developing web-based demo ⚛️ on marking medical images with NLP 🤖
📝 Paper card: nicolayr.com#casa2025
#ManchesterHealthSeries #Healthcare
