AI Tech Desk for Multifaceted Problems
Budget: $750 – $1,500 USD
I'm in need of an AI-Powered Technical Service Desk that can effectively handle a variety of technical issues, particularly those related to hardware, software, and network setups.
Overview:
I'm looking to create an MVP of an AI-assisted technical service desk tool. The goal is to prototype the core capabilities of ingesting, analyzing, and responding to support issues using machine learning techniques.
Project Description:
The tool will allow end users to upload tickets, logs, configs, and other pertinent files. It will parse, extract, and vectorize content to understand issues. An explainable ML model like LIME will then diagnose the problem and provide steps, code samples, etc to resolve it. The LLM will be RAG augmented based on a vector protocol of a list of sites to augment the LLM.
Key Criteria for Success:
Robust technical execution that meets core requirements
Modular, extensible architecture for flexibility over time
Quality, well-tested code delivered on time
Clear, prompt communication throughout project
Key Components:
Text/file ingestion and vectorization
Semantic similarity search in Milvus/SPTAG
Local post-hoc model explanations
Production-ready Flask/FastAPI UI
Documentation and usage guides
Document Ingestion (spaCy, Beautiful Soup, PyMuPDF): Scrape/ingest docs, extract text, handle formats
Vectorization (Sentence/Doc2Vec): Convert text to distributed representations
Vector Database (Milvus, SPTAG): Store/index vectors at scale for retrieval
Semantic Search (FAISS): Leverage ANN for related content finding
Knowledge Graph (spaCy, NetworkX): Extract entities/relations, build graph
Explainability (AIX360, LIME): Provide post-hoc model explanations
UI (Flask, Streamlit): Build search/browse front-end for end users
This is a fixed price project. Do not ask to chat or set meetings to discuss requirements. Bids will not be binding.
Scoring: First selection will be the top 5 lowest bidders. Second selection will be answering any questions that are asked. Anyone who submits a new bid that is more than 10% of their initial bid will be eliminated due to not doing their due diligence. Final selection will be made on final proposal (time/cost/incentives). If the winner does not perform or communicate, the project will be cancelled and the next bidder will be asked to come in. Payment based on final milestone delivery only. Do not ask for hourly or multiple milestones - I will not accept those bids.
Overview:
I'm looking to create an MVP of an AI-assisted technical service desk tool. The goal is to prototype the core capabilities of ingesting, analyzing, and responding to support issues using machine learning techniques.
Project Description:
The tool will allow end users to upload tickets, logs, configs, and other pertinent files. It will parse, extract, and vectorize content to understand issues. An explainable ML model like LIME will then diagnose the problem and provide steps, code samples, etc to resolve it. The LLM will be RAG augmented based on a vector protocol of a list of sites to augment the LLM.
Key Criteria for Success:
Robust technical execution that meets core requirements
Modular, extensible architecture for flexibility over time
Quality, well-tested code delivered on time
Clear, prompt communication throughout project
Key Components:
Text/file ingestion and vectorization
Semantic similarity search in Milvus/SPTAG
Local post-hoc model explanations
Production-ready Flask/FastAPI UI
Documentation and usage guides
Document Ingestion (spaCy, Beautiful Soup, PyMuPDF): Scrape/ingest docs, extract text, handle formats
Vectorization (Sentence/Doc2Vec): Convert text to distributed representations
Vector Database (Milvus, SPTAG): Store/index vectors at scale for retrieval
Semantic Search (FAISS): Leverage ANN for related content finding
Knowledge Graph (spaCy, NetworkX): Extract entities/relations, build graph
Explainability (AIX360, LIME): Provide post-hoc model explanations
UI (Flask, Streamlit): Build search/browse front-end for end users
This is a fixed price project. Do not ask to chat or set meetings to discuss requirements. Bids will not be binding.
Scoring: First selection will be the top 5 lowest bidders. Second selection will be answering any questions that are asked. Anyone who submits a new bid that is more than 10% of their initial bid will be eliminated due to not doing their due diligence. Final selection will be made on final proposal (time/cost/incentives). If the winner does not perform or communicate, the project will be cancelled and the next bidder will be asked to come in. Payment based on final milestone delivery only. Do not ask for hourly or multiple milestones - I will not accept those bids.