Quantiphi, Inc

Disclosed is computer-implemented method for automating document analysis process. Method comprises: receiving plurality of documents into document classification module; processing each document using document classifier to classify each document into related document (RD) or unrelated document; when given document is classified as RD: segmenting given document into plurality of data chunks (PDC) comprising text portions, generating semantic embeddings (SE) for each data chunk of PDC, indexing PDC and corresponding SE thereof, in database, retrieving SE of PDC for summarizing the given document based on: document identifier, subject, generating subject-based query embedding, performing similarity search of subject-based query embedding to retrieve SE of data chunk related to subject, and passing retrieved SE to large language model (LLM) to generate summary of given document.

A system ( 100 ) for fine-tuning a code translation model is disclosed. The system comprises a processing arrangement ( 102 ) and a user device ( 104 ). The processing arrangement is configured to receive at least one seed sample in a source code language, a code translation base model, and a set of system specific requirements, from a user device. The processing arrangement is configured to generate synthetic dataset comprising training dataset and evaluation dataset; determine at least one parameter to finetune the code translation model; finetune the code translation model based on at least one parameter; evaluate the code translation model to generate a evaluation score; determine a training loss and an evaluation loss; based on the evaluation score, the training loss and the evaluation loss, finetune the code translation model or approve the code translation model. A method for fine-tuning a code translation model is also disclosed.

System ( 100 ) for optimization of industrial operation, based on simulation data thereof, system comprising processing unit ( 102 ) configured to generate simulation data related to industrial operation; generate knowledge graph ontology; generate knowledge graph instance representing simulation data; receive user request to analyze quality of industrial operation, from user device ( 104 ) of user; analyze quality of industrial operation; generate response to user request, based on analyzed quality of industrial operation; send response to user device; receive first user input from user device, wherein first user input is indicative of user approval on industrial operation; and when user approval on industrial operation is positive, employ simulation data for implementation of industrial operation; or when user approval on industrial operation is negative, generate updated simulation data, based on analysis of quality of industrial operation, and repeat aforementioned steps for updated simulation data.

A method and system for evaluating Artificial Intelligence (AI) agents is disclosed. The method includes receiving a profile of the AI agent. The method may further include generating synthetic data based on the received profile of the AI agent. The method may further include interacting with the AI agent to generate traces when the AI agent is interactive. Further, the method includes generating an evaluation score based on the traces of the AI agent with respect to a corresponding evaluation metric by a plurality of evaluator agents. The method further includes generating an evaluation report by consolidating the evaluation scores of each of the plurality of evaluator agents.

Disclosed is method for distributed decision-making in multi-role large language models (LLMs) architecture (mLLMa). Method comprises: receiving user query for initiating conversation between user and mLLMa; using graphical representation (GR) for identifying role(s) associated with user query, wherein role(s) is based on context thereof; assigning relevance score to each role; dynamically passing role(s), to mLLMa; generating role-specific prompt for role assumed by LLM; generating, by each LLM, role-specific response (RR) corresponding to role-specific prompt for each role and context; presenting RR from each LLM to peer LLMs; conducting polling process among LLMs for ranking RR therefrom; aggregating rankings from polling process to determine final ranking of RR for each LLM; selecting RR having highest final ranking as an action-inducing response (AR), and transmitting AR to user for providing user action; and updating GR based on AR and user action.

Disclosed is method for automated code retrieval and code generation, comprising: receiving code repository ( 302 ); extracting coding components (CC); generating knowledge graph (KG) comprising nodes representing extracted CC, and edges representing relationships between the extracted CC; identifying node(s) in generated KG for indexing; creating first index (FI) ( 406 , 512 ) and second index (SI) ( 408 , 514 ) of generated KG; receiving user request ( 502 ) for generating code; searching FI and SI based on entity(s) ( 510 ) extracted from user query (UQ) and embeddings ( 506 ) of UQ, respectively, for identifying predefined number of nodes (PNN) ( 516 , 520 A) in KG; retrieving identified PNN from KG; retrieving neighbouring nodes (NN) ( 520 B-C); generating subgraph of retrieved NN and relationships between retrieved NN; filtering generated subgraph (GS); retrieving code snippets ( 404 C) of retrieved NN; generating contextualized subgraph; and prompting Large Language Model ( 608 ) with generated contextualized subgraph, predefined KG ontology and area of interest(s), for generating code ( 610 ).

A method ( 400 ) and system ( 100 ) to prepare a framework for enzyme optimization is disclosed. The method ( 400 ) includes executing a plurality of optimization modules to obtain engineered enzyme variants. The method ( 400 ) may include conducting a similarity search against one or more known enzyme databases. The method ( 400 ) may further include obtaining a multi-property optimization (MPO) score by evaluating the engineered enzyme variants generated by the optimization modules according to predefined goals. The method ( 400 ) may further include generating a feedback based on the MPO score to iteratively refine the optimization modules and improve enzyme variant properties.

Disclosed is method for generating a target molecule ( 412, 514, 802 ), comprising receiving first user input ( 806 ) indicative of properties associated with target molecule, and identifying properties ( 808 A-C) associated with targeted molecule and corresponding objectives ( 810 A-B); generating property scores ( 832 A-B) for properties using property predictor algorithm ( 812 ); receiving second user input indicative of molecular structure ( 202, 402, 502, 602, 702, 814 ) of input molecule ( 204 ); generating corresponding target molecules (CTMs) ( 200, 406, 508, 600, 700, 816 ); generating embeddings ( 824 ) of CTMs; determining aggregate similarity score ( 828 ); determining aggregate property score; determining fitness scores ( 410, 512, 834 ) of CTMs; determining whether given target molecule amongst CTMs fulfill termination criteria (TC); when it is determined that TC is fulfilled by given target molecule, deeming given target molecule as target molecule to be generated; when it is determined that TC is not fulfilled, updating generated CTMs, iteratively performing steps (v) to (ix).

Disclosed is system ( 100 ) for generating visualizations from structured data (SD) queried through natural language. The system comprising processor ( 102 ) communicably coupled to user device (UD) ( 104 ), configured to: receive partially unstructured data query from UD and generate corresponding SD query; execute generated structured data query (SDQ) to generate SD comprising one or more datapoints; receive and analyze SD according to one statistical principle; when analyzed SD comprises plurality of datapoints (POD), generate visualization recommendation (VR) based on generated POD; or when analyzed SD comprises single datapoint, generate at least one alternative SDQ corresponding to one of: partially unstructured data query, SD query, execute generated at least one alternative SDQ to generate alternative SD comprising one or more datapoints, analyze alternative SD according to one statistical principle, and when analyzed alternative SD comprises POD, generate VR based on generated POD; and present, at UD, generated VR to user.