Term
| What AI based projects have you worked on? CLEARED4 Proactive ID Monitoring - BLUEDOT |
|
Definition
**Using BlueDot's AI - draw on global disease surveillance data / predictive modeling to anticipate outbreaks
**C4 helps Uni or Corp develop targeted response plan including preemptive health screenings, vax drives, tailored comm's
**C4 provides realtime updates on outbreak's progression - enable uni to adapt response as needed
**RESULT? Lower infection rates, minimal disruption to academic operations, well informed student population that feels safe |
|
|
Term
| How does BlueDot AI work? |
|
Definition
| Employs NLP and ML to sift through collected data every 15 min - 24hrs - AWS..By cross referencing disease data w additional data sets such as air-traffic iteineraries, BD can predict risk of outbreak spreading / identify regions effected next. |
|
|
Term
| How does C4 use Bluedots specifically? |
|
Definition
We use their Data as a Service model - API's to have access to BD's infectious disease surveillance data - real time data on infectious disease for predictive modeling, risk assessment
We use this dataset in our pro tier pricing model for ID Risk Analysis to provide
1. Real Time Health Monitoring/Alerts:
Data on new cases, recovered, active - visual indicators - color coded maps, graphs..access control and geofencing data based on locale. Alerts to admin and staff.
2. Customizable health surveys - track symptoms/exposure among students/staff -
3. Contact tracing integration - integrates CT capability to help ID and notify individuals who have been in CLOSE CONTACT, CONTACT, file as HIGH RISK, MEDIUM, LOW - Quarantine
4. Predictive analytics - ability to provide a RA score 1 to 100 based on student risk makeup, state and local risk
5. Configurable rulesets automation - after major breakout, before, during, |
|
|
Term
|
Definition
SURPLUS LINES TAX FILING, COMPLIANCE SOLUTIONS, INSCIPHER CONNECT
1. Surplus Lines Tax Filing:
Provide services/software for insurance professionals to simplify process of surplus lines tax filing / reporting.
2. Compliance Solutions:
offer solutions to manage compliance requirements that vary by state, including automated surplus lines tax calculators/API's
3. Software Dev - InsCipher Connect - intuitive surplus lines tax mgmt software - streamlines compliance process for SL tax rates / deadlines |
|
|
Term
| What is surplus lines tax filing? SLTF |
|
Definition
| SL tax filing is process where insurance agents / brokers pay a state mandated tax for policies they place with non-admitted, specialized insurers |
|
|
Term
|
Definition
SaaS that facilitates agents and carriers with the handling of surplus lines business
Traditional insurance carriers are subject to certain rules / financial regs. they must adhere to do business in each state
Surplus line insurers are not subject to these regulations - so the premium in SLI is taxable. With this taxable premium many rules/regulations on the state level as to what is required to be compliant with surplus lines filing of those taxes - like: Varying state tax pct, Stamping fees related to surplus lines, forms that insureds need to sign, deadlines on premium tax
Because of all these rules/regs it becomes difficult for an agent or carrier to file or handle business in all 50 states. Inscipher provides service to untangle web of rules/regs |
|
|
Term
| Who are Inscipher's clients? |
|
Definition
1. Retail agents
Ins. agents who directly sell ins. products to end customer
2. Wholesale brokers
Intermediaries who facilitate placement of ins. policies between retail agents / insurers - especially in SL market
3. MGA's (Managing General Agents)
Firms/individuals authorized by insurer to manage all or part of insurer's biz activities, including underwriting, pricing, appointing retail agents. |
|
|
Term
| How is SLI different than traditional insurance? |
|
Definition
Mainly in the risks it covers and regulatory env. its in
1. Covers nonstandard risks: SLI covers for unique or high risk situations that standard insurers avoid, like insuring rare art collection or business w/ high liability risk
2. Regulatory Flexibility: traditional ins. is heavily regulated by state ins. dept's, SLI operates with more flexibility, allowing it to offer customized policies for unconventional risks.
3. Market of Last resort: for businesses / individuals who cannot obtain coverage through standard markets due to nature of risk |
|
|
Term
| Health Credential intake / Nanonets OCR - tell about it |
|
Definition
We used Nanonets OCR to:
<> DATA EXTRACTION: Automatically extract key info such as name, test result, vaccine type, date of administration from doc.
<> VALIDATION / INTEGRATION:
Extracted data is validated thru NN AI models to ensure accuracy. Once validated, info is integrated onto dashboard - status
<> Access control
|
|
|
Term
| Explain Nanonets / how you trained |
|
Definition
1. Credential validation model: Nanonets OCR - used to automate workflows through AI by interpret/recognize text from images/scanned docs. Converts types of docs into machine readable data - invoices, receipts, DL's, vax, test result, passport - extracts data in K/V pair.
How did we train?
Specific to health credential space we tested increasing # of images of vaxxcards, test results, ID's. NN has continuous learning ability where model improves accuracy/understanding w each new doc processed.
STEPS
1. GET THE REPO: Clone GH repo, get API key, set key env
2. CREATE MODEL: Create a new model - use python script to create new OCR model
3. UPLOAD TRAINING DATA/TRAIN: Upload training data, train the model..involves model learning to recognize/extract text
4. MAKE PREDICTIONS submit new docs to model and receiving data in return
HOW TO TRAIN MODEL (CLUT) - COLLECT, LABEL, UPLOAD, TRAIN
Collect/prep training data - various formats, layouts model will encounter
Label data - annotate training data by marking areas you want to extract - specify keyvalue pairs or fields you want
Upload labeled dataset - via API
System use deep learning algorithms to learn the labeled data understanding how to ID and extract specified info from similar docs |
|
|
Term
| EXPLAIN HIGH LEVEL WHAT AI you have worked on |
|
Definition
HALDA.AI
Hal -> AI chatbot:
AI conversations, AI recommendations
Collect data we want it to learn from, train it, run script that feeds data into CGPT model - scrape datasets, dept of ed
CLEARED4
Health credential validation -> Nanonets OCR AI
Used NN OCR AI to extract data based on trained models we created to look for labels, validate based on looking at examples of other docs similar
Risk analysis -> Bluedot
structured data assets, forecasting models, AI surveillance engine -> we used their API to gather data on ID's and anticipate surges in virulence, positivity, etc. --> Real time alerts for customers, Suggestions, Configurable playbooks, Risk assessment model,
Raffle.AI - NL
Semantic Search -> Raffle.AI
Natural language search (NLP) to understand interpret queries in humanlike manner. Allows search using natural language - no keywords. Semantic search understand meaning behind queries to provide accurate answers - achievd thru NLP algo's
HOW TO TRAIN? Used pretrained model like BERT
|
|
|
