Sensefier

and

AESTIMA SUPER BOT

LLM enabled tool for conducting research

Latest news
Learn More
↓

SENSEFIER

Contact us Support bot info@sensefier.com
The bot aggregates expert opinions on the price of the selected securities and makes these insights accessible to our users. We deal with opinions that are publicly expressed by experts. Our service does not evaluate or synthesize these opinions into a consensus view; rather, we present them as they are, directly reflecting the perspectives of the experts. LLM determines if the source contains an expert opinion. LLM assesses the source and decides if it is sufficient to formulate a structured viewpoint. More info

Superbot

1. Aim

The aim of the AESTIMA Superbot is to improve, with the use of AI, evidence analysis and synthesis for research, education, policy making or commercial research and development purposes. The AESTIMA Superbot represents a significant advancement in the field of AI-powered evidence analysis and synthesis. By integrating cutting-edge technologies like ChatGPT, LangChain, Zotero, and PineCone, it offers a powerful and efficient tool for researchers and decision-makers across various fields. As AI continues to evolve, tools like the AESTIMA Superbot will play an increasingly important role in shaping the future of research and decision-making.

2. Strategic Objectives

Creation of domain agnostic bots empowered by LLM(s) and pre-designed set of data management tools, to facilitate seamless integration with user’s knowledge base, user’s search trajectory.

3. Priorities

  1. Addressing digital transformation through development of digital readiness, resilience & capacity to the research process
  2. Particular attention will be given to promoting gender and other equalities & addressing differences in relation to the access & use by groups with limited access to specialized expertise and/or research resources.
  3. Inclusion & diversity in all fields of education, training: the tool will support projects that promote development of projects relevant for social and economic development.
  4. Creation & implementation of internationalization strategies: The tool will be available regardless of user’s location. Clear guidance on the tool use in multiple languages will be provided.

4. Product Technical Description

The AESTIMA SuperBot operates on a sophisticated technical framework that integrates several advanced technologies. At its core, it uses ChatGPT, a state-of-the-art language model developed by OpenAI. This technology enables the SuperBot to generate human-like text based on the input it receives, allowing it to interact intelligently with users. In addition to ChatGPT, the Superbot also utilizes LangChain, a Python-based language processing library. LangChain aids in the processing and understanding of natural language, enabling the Superbot to comprehend and respond to complex queries effectively. The Superbot's functionality is further enhanced by its integration with Zotero, a free, easy-to-use tool that helps users collect, organize, cite, and share research. In its current version, the Superbot receives a database of PDF documents, such as articles or books stored in Zotero. It then performs vectorization on these documents, a process that converts the text data into a numerical format that can be processed by machine learning algorithms.

Proposals for further research and development

diagram

Research tools integration

Development directions – collaboration opportunities
  1. Full automation and integration of data search, extraction and relevance check process.
  2. Possible advancing in data vectorization and similarity search.
  3. Full automation and integration of data extraction process (text only).
  4. Testing alternative LLMs, enhancing context and finding specialization.
  5. Integrating specialized filters and path tracks such as ABS check, PRISMA, source quality assessment, docs selection etc.
  6. Integrating user’s ability to create and manage own knowledge base (Zotero and other).
  7. Integration with data coding and labeling tools (MaxQDA) and developing semantic coding capabilities.
  8. Quality assessment tools for the sources included in literature reviews

Quality of response

Development directions – collaboration opportunities
  1. Enhancing Reliability and Trustworthiness
    • Verification Mechanisms: Solicit ideas and contributions for building robust verification mechanisms that can automatically check the accuracy of the information provided by the LLM. This could involve cross-referencing with trusted databases or incorporating a layer of peer review.
    • Hallucination Detection Features: Focus on developing advanced algorithms or methods to detect and flag potentially inaccurate or "hallucinated" information generated by the LLM.
  2. Optimizing Model Finetuning and Adaptation
    • Finetuning for Specialized Domains: Request contributions for strategies or methodologies to finetune the LLM on specialized datasets without compromising the model's generalizability or introducing biases.​
    • Customization and User-Driven Learning: Explore ways to allow users to contribute to the model’s learning process, tailoring it to better suit their specific research needs and domains.
  3. Semantic Analysis and Text Handling Improvements
    • Advanced Text Analysis: Since direct word and symbol counting is a challenge for LLMs, seek contributions for innovative methods or workarounds that can achieve similar outcomes, such as estimating text length or complexity through semantic analysis.​
    • Enhancing Textual Understanding: Encourage the development of features or plugins that improve the LLM’s ability to understand and process user queries, especially for complex academic texts, using advanced natural language processing techniques.​
  4. Alignment and Ethical Use​
    • Alignment with User Intent: Ask for input on creating more sophisticated mechanisms to ensure the LLM’s responses are better aligned with user intentions, especially in the context of research, where precision and relevance are critical.​
    • Ethical Guidelines and Use Cases: Engage the community in establishing clear ethical guidelines for the tool’s use, ensuring it promotes integrity in research and respects intellectual property.​
  5. Legal restrictions​
    • Investigation and adoption of key AI regulations. Following the fact that we are not designing and training LLMs, but allow user to create new content with help of LLMs. ​

Our team

Alex Filatov
Co-Founder & Investor, EMBA
Since 2018 Alexander is a co-founder and CEO at EverX – core developer of the Everscale blockchain platform. He is an entrepreneur and former executive with 28 years of experience at leadership positions in marketing, business development and general management.
Dmitrii Gimmelberg
Co-Founder & Investor, MBA
Dmitrii is an experienced CEO, CFO, product and business development manager with 29 years of experience, working across several countries and industries, including investments, industrial digital solutions, retail and service. Dmitrii is PhD student, his scientific interest is in LLMs disruption of businesses.
Marta Głowacka
Lead Researcher, PhD
Marta holds a PhD in Health Psychology Research and Professional Practice and MSc Health Psychology and has 15 years of research experience working in multidisciplinary teams. She also has experience teaching research methodology and psychology at university level. Her interests include research methodology, especially remote methods, evidence review and synthesis, and environmental health psychology.
Alexei Belinskiy
Investor and supporter, DBA
Alexei has 28 years of experience in entrepreneurship, investment, digital technologies and C-level management. He holds a MSc degree in Physics and Doctor of Business Administration studies. His scientific interest are in the field of underlying causes of managerial success.
Sergey Korotkiy
Tech Lead, PhD
Sergey is an experienced product manager with 20 years of experience in large international IT companies. Ph.D. of Engineering Sciences. B2B technology entrepreneur & educator.
Valentin Artamonov
Lead Developer
Valentin has 15 years of experience in AI, application development and testing using C# and Python. Throughout his career, Valentin has contributed significantly to developing key software products for two leading banks and a robotics company.

Contact with us if you are interested in

joint research
product commercialisation
fundraising