INFINYTY.TECH

• Challenge:
• Manual hiring workflows were bottlenecked. Creating role-specific Google Forms took 4 hours, and manual candidate fit reporting required 2 hours per applicant.

• Solution:

       ✔ AI Question Engine: Render-based UI 

      accepts inputs like industry, role,    

        seniority. It auto-generates tailored 

      evaluation Qs and programmat-   

    ically builds Google Forms.             

       ✔ AI-based Proprietary Fit Assessment:

     AI workflow evaluates candidate   

             responses and generates user-facing 

         reports for HR instantly. Auto-emails

report to HR  .                                  

• Impact:

        ⮚ Question generation, form creation         

       and report generation time slashed  

from 4+ hours to under 2 min      .

   ⮚ Enabled client to handle significantly

           higher volume of roles and candidates

without increasing headcount     

• Challenge:
• Online recruitment platforms struggled with candidate fraud, specifically screen-reading or using generative AI during live interviews.

• Solution:

        ✔ Multimodal Analysis: Engineered a     

          proprietary AI algorithm to analyze   

          synchronized video & audio feeds for

               micro-behaviors, including saccadic       

              eye movements and specific

facial tilts

indicative of screen reading.     

      ✔ Plagiarism Check: Converted voice-  

              to-text, and checked for AI-generated  

     responses or plagiarized answers.

• Impact:

      ⮚ Significantly increased  the precision 

     of fraud flagging .                              

• Challenge:
• Manual synthesis of 10,000+ pages of research into 200- page FDA dossiers took 2 experts ~4 months, and was highly prone to clerical and cross-referencing errors.

• Solution:

   ✔ Intelligent Document Synthesis:   

  Developed AI pipeline (LLMs)  

               to ingest massive datasets, generating   

   structured approval sections with

automated  cross-referencing  

       ✔ Human - in - the - Loop UI:  Built   custom     

          interface for modular review, allowing 

         experts to manually refine sensitive      

       sections  while automating standard technical data                                  

• Impact:

        ⮚ Reduced end-to-end production time

             from 16 weeks (and 2 experts) to just  

 2 weeks  with a single operator

⮚ Achieved   95 %   automated         

     generation, with only 5% requiring

 direct  human intervention.       

• Challenge:
• Dealerships were overwhelmed with leads, many of which were low quality. Sales teams lacked a reliable way to prioritize outreach and maximize conversions.

• Solution:

✔ Built data science model to score

      incoming leads basis historical      

conversion data, credit profile, 

location, and vehicle type and 

               integrated it into CRM real-time.                  

    ✔ Developed a voice-based AI agent 

           (powered by LLM and telephony API) 

               that automatically called high-scoring   

   leads and scheduled follow-ups

• Impact:

     ⮚  40%  reduction in manual cold calls    

       made by dealership teams.               

       ⮚  Increased revenue per dealership by  

18% over 3 months.                        

• Challenge:
• Dealerships were overwhelmed with leads, many of which were low quality. Sales teams lacked a reliable way to prioritize outreach and maximize conversions.

• Solution:

✔ Built data science model to score

 incoming leads basis historical 

  conversion data, credit profiles, 

 location, and vehicle type and 

    integrated it into CRM real-time

    ✔ Developed a voice-based AI agent 

            powered by LLM and telephony API) 

            that automatically called high-scoring  

  leads and scheduled follow-ups

• Impact:

     ⮚ 40 % reduction in  manual cold calls

made by dealership teams      

          ⮚  Increased revenue per dealership by   

18% over 3 months                      

• Challenge:
• Siloed data caused difficulty attributing revenue to ad spend, and budget allocation challenges due to partial insights

• Solution:

        ✔ Created a unified data model using    

        product, marketing and acquisition 

 data.                                    

✔ Built interactive dashboards  in 

           Superset for ROAS and LTV tracking

        ✔ Provided insights on high performing   

          marketing channels and campaigns

• Impact:

    ⮚ 22% reduction in wasted ad spend in 

the first month                                 

         ⮚ 1.3x improvement in blended ROAS after

6 weeks                                             

      ➢ Real- time ROAS reporting adopted   

     by marketing leadership, enabled

responsive budget planning     

• Challenge:
•  ✔ Built unified customer profiling, cross-sell propensityarketing spends across channels/customers, leading to low cross-sell conversion rates across personal loan and two-wheeler loan.

• Solution:

        ✔ Built unified customer profiling, cross-   

           sell propensity modeling, and enabled

multi-channel personalized         

marketing ( Email, WhatsApp,     

CleverTap)                                       

        ✔ Implemented a marketing attribution  

           model to track cross -channel impact and ROI dashboards                    

• Impact:

⮚ 18% increase  in loan cross - sell

                conversions within 2months                      

        ⮚ 15% reduction in cost per acquisition 

             (CPA) by optimizing underperforming 

  channels