Founders' Adda: Raw Conversations with India’s Top AI Pioneers

Summit Talk Summary

---

Key Topics Covered

• Industrial Automation & Robotics — Using agentic AI to democratize factory automation
• Voice Infrastructure for India — Call center automation and multilingual voice agents
• Geospatial Intelligence — Real-time map updates using video analytics
• Healthcare AI — AI-driven cancer diagnosis and radiation therapy planning
• Voice Architecture & Natural Language Processing — Emotion-aware speech systems and Indian dialect mastery
• Data Privacy & Compliance — DPDP (Data Protection), HIPAA, ISO certifications, data sovereignty
• Scaling Challenges — From demos to production, handling concurrency, quality at scale
• Foundational Models vs. Application Layers — When to build custom infrastructure vs. leverage existing models
• Indian Language & Dialect Diversity — Solving problems unique to India's linguistic landscape
• Cost Optimization — Reducing operational expenses through AI automation (70%+ in some cases)

---

Key Points & Insights

1. Agentic AI for Industrial Automation (Technoate AI)

   • Moving from DIY robotics conceptualization to deployment to troubleshooting using AI agents
   • Challenges in India: capital constraints prevent building foundational models; approach is customer-first experimentation
   • Robotics requires real-world data injection; LLMs alone (ChatGPT) cannot handle domain-specific robot programming and CNC programming
   • Deployment with Fortune 500 companies and Indian Air Force validates problem-solution fit

2. Call Center Automation at Scale (Quantis AI)

   • India's $55B customer support industry can be fully automated with voice agents handling listen-understand-act-respond cycles
   • Custom data engines critical to overcome public dataset limitations; built proprietary data generation infrastructure
   • Concurrency scaling: currently ~50 concurrent calls; scaling to enterprise demand (e.g., SBI, insurance companies)
   • Pricing model: per-minute subscription; achieves 90% cost savings for large enterprises (₹3/min vs. ₹25-30/min human agent)
   • Real-world deployment: real estate outbound calls, lead qualification, site visit booking—full end-to-end automation

3. Video Analytics & Real-Time Mapping (Papy Labs)

   • Processing 100+ petabytes of video data from vehicles and CCTV across Indian cities
   • Use cases: dynamic billboard pricing (40-45% revenue increase for JC Decaux), autonomous vehicle safety, bus fleet optimization, news generation
   • Data privacy approach: blur faces and license plates, use bare metal (not cloud hyperscalers), store in Europe for GDPR compliance
   • Key insight: India has data generation capability but needs infrastructure for permissioned data use without hyperscaler dependency

4. AI-Driven Cancer Diagnosis & Treatment Planning (Imagics AI)

   • 20M new cancer cases annually; bottleneck is shortage of oncology experts, not diagnostic capability
   • Solved: automatic organ contouring for radiation therapy planning (reduced from 60-90 min to 5-15 min)
   • Dataset: 5M images, 30% from Northeast India (collected on-premise with NITI Aayog support)
   • Human-in-the-loop design: AI assists but radiologist approves; not replacing doctors
   • Certifications: HIPAA, ISO 13485, CDSCO licensed (mandatory for medical device software in India)
   • Real-world impact: 1M CXR scans analyzed, 4,000+ TB cases detected, 6 early-stage lung cancers identified

5. Indian Dialect-Native Voice Infrastructure (Industs AI)

   • Global players (Google, 11Labs) offer generic Hindi; India's dialects change every 20km
   • Built end-to-end stack: Speech-to-Text (S2T), Text-to-Speech (TTS), LLM, Speech-to-Speech, all optimized for Indian languages
   • Recently launched emotion-aware S2T model; recognizes laughter, anger, sentiment in real time
   • Latency: 300-400ms (vs. 11Labs ~500ms); Cost: ₹2/min vs. ₹8/min (11Labs)
   • Data residency: all data stays in India (sovereign cloud infrastructure)
   • Platform: DIY no-code agent builder; integrations with banks, FMCG, enterprises; API-based for developers
   • Partners: Airtel, Jio for SIP channel integration; expanding to 5+ global languages (Arabic, German, French, Mandarin)

6. Foundational Models vs. Application-Layer Strategy

   • Technoate AI: Started wanting to build foundational model; after customer feedback, realized 3D real-world data + industrial domain specificity requires custom infrastructure. Complex workflows > simple chat responses.
   • Quantis AI: Public datasets (Bhashini, Google) failed at scale; built custom data engine. Each use case needs fine-tuning.
   • Industs AI: Pivoted from using third-party TTS → discovered third-party limitations (pronunciation, dialects) → built proprietary infrastructure.
   • Insight: Foundational models are necessary but insufficient; application-layer customization is where defensibility lies.

7. Data Generation & Quality as Competitive Advantage

   • Quantis AI: Won Prime Minister Modi award for data generation engine
   • Papy Labs: 100+ petabytes of video require annotation and categorization; search capability critical
   • Imagics AI: 30% Indian data in training crucial for Northeast India accuracy
   • Theme: In India, permissioned, proprietary, use-case-specific data > public datasets

8. Compliance & Trust in Regulated Sectors

   • Healthcare: HIPAA, ISO 13485, CDSCO certification; human-in-the-loop mandatory for radiologist approval
   • Geospatial/Privacy: DPDP Act; blur PII, use bare metal, avoid hyperscalers, data residency in Europe for GDPR
   • Voice: Data sovereignty (India-based servers); no data sharing with global hyperscalers
   • Insight: Compliance is not a checkbox—it's a product feature and differentiator in India

9. Cost Economics & ROI

   • Call centers: 90% cost reduction for large enterprises (₹3/min AI vs. ₹25-30/min human)
   • Voice infrastructure: 70% cost reduction (₹2/min vs. ₹8/min competitors)
   • Fleet optimization (DTC): Reduced revenue loss by improving passenger-to-capacity ratio; saved on operational inefficiencies
   • Entry pricing: Geospatial tiles (Papy Labs) ₹1.5L/25km²/day; scales with volume

10. Market Segmentation & Pricing Reality

• Small companies (5-10 employees): not yet viable; pricing will decrease in "couple of years"
• Current focus: Enterprise (SBI, insurance, 10K+ employees with high per-minute contact volume)
• Long-term TAM: As pricing drops, SME adoption will unlock

---

Notable Quotes or Statements

• Davindar (Technoate AI): "We are aiming to automate automation itself... AI won't take away any jobs? Let us do something about it."

• Webhaward Shukla (Quantis AI): "India is the customer support capital of the world... the entire model is outdated in the agent era."

• Praum (Papy Labs): "The problem in all billboards is they come at standard price... we brought a new pricing mechanism based on impression count and increased revenue 40-45%."

• Minel Gupta (Imagics AI): "There are around 20 million new cancer cases every year... the main problem is shortage of clinical experts... cost lives or changes the stage of the cancer."

• Vive Gupta (Industs AI): "After each 20 kilometer dialect changes... we need a company in India who can build the infrastructure of voice in our country based on our directs."

• Vive Gupta (Industs AI): "If somebody's laughing over the call how would AI system recognize the person is happy or angry? That's how the agent would say sorry or congratulate because you need to understand emotions."

• Moderator (opening remarks): "The way... the best way to learn at a conference like this is by listening to each other... it's only about product. No business, no funding conversation."

---

Speakers & Organizations Mentioned

Founders & Companies

• Davindar Kumar — Technoate AI (robotics automation)
• Webhaward Shukla — Quantis AI (voice infrastructure, call centers)
• Praum Gupta — Papy Labs (geospatial video analytics)
• Minel Gupta, Fatma, Sheal Taras — Imagics AI (cancer diagnosis & radiotherapy planning; "three Divas" per Modi)
• Vive Gupta — Industs AI (voice OS for Indian languages)

Government & Institutions

• Narendra Modi (Prime Minister) — recognized Imagics AI founders and Quantis AI data engine
• Bill Gates — showed Imagics AI solution to him
• NITI Aayog — supported data collection in Northeast India
• Indian Air Force — deployment use case for Technoate AI
• CDSCO (Central Drugs Standard Control Organization) — medical device licensing
• ICMR (Indian Council of Medical Research) — certifies software as medical device

Corporate Partners

• Fortune 500 companies — early deployments (Technoate AI)
• PTM (likely "Parcel Tracking/Logistics") — Quantis AI customer at scale
• OpenAI — partnership with Quantis AI for infrastructure
• JC Decaux — billboard company, Papy Labs customer (40-45% revenue increase)
• MG Motors / Hector — autonomous rental, Papy Labs customer
• BCG — consulting firm, DTC route rationalization, Papy Labs customer
• Delhi Transport Corporation (DTC) — 8,000 buses, funded by JICA (Japan International Cooperation Agency)
• SBI, Insurance companies — Industs AI customers
• FMCG enterprises — Industs AI customers
• Airtel, Jio — SIP integration partners for Industs AI
• Bhashini, Google, 11Labs — referenced as benchmarks (not partners)

Award Bodies

• Prime Minister Modi — recognized Imagics AI and Quantis AI

Conferences & Venues

• Bharat Mandapam — venue of the summit (New Delhi)

---

Technical Concepts & Resources

AI/ML Models & Architectures

• Agentic AI — Used by Technoate AI for autonomous decision-making in robotics
• Foundational Models — Technoate AI, Quantis AI, Industs AI each built custom FMs or pivoted to app-layer strategies
• Generative Vision / Generative VI — Robot programming, CNC programming, visual error diagnostics (Technoate AI)
• LLM — Speech-to-Text, Text-to-Speech, Speech-to-Speech integration (Industs AI)
• Fine-tuning — Overcoming public dataset limitations (Quantis AI, Imagics AI)
• Emotion-Aware Speech-to-Text — Recently launched model recognizing laughter, sentiment (Industs AI, launched at Bharat Mandapam)

Data & Privacy Technologies

• DPDP Act (Digital Personal Data Protection) — India's privacy law; blur PII (Papy Labs)
• HIPAA — Healthcare compliance (Imagics AI)
• ISO 13485 — Medical device certification (Imagics AI)
• CDSCO License — Required for medical device software commercialization in India (Imagics AI)
• Bare Metal Servers — Infrastructure choice to avoid hyperscaler lock-in and ensure data residency (Papy Labs, Industs AI)
• GDPR — European data residency compliance (Papy Labs stores in Europe)
• Data Engines — Proprietary data generation, annotation, categorization systems (Quantis AI, Papy Labs, Imagics AI)

Imaging & Medical

• CT Scan, MRI — Input data for radiotherapy planning (Imagics AI)
• DICOM — Standard format for medical imaging; custom DICOM viewer built (Imagics AI)
• Contouring — Segmentation of organs at risk around tumors (Imagics AI)
• CXR (Chest X-Ray) — Used for TB and lung cancer screening (Imagics AI; deployed across Gujarat, 1M+ scans)
• Radiotherapy Planning — Automating dose optimization and organ sparing (Imagics AI)

Voice & Speech

• S2T (Speech-to-Text) — Sub-400ms latency optimization (Industs AI)
• TTS (Text-to-Speech) — Multilingual support (Industs AI, Quantis AI)
• Speech-to-Speech — Direct speech translation (Industs AI)
• Sentiment Analysis — Real-time call sentiment extraction (Quantis AI, Industs AI)
• Voice Cloning — Custom voice synthesis on platform (Industs AI)
• Indic Languages — Hindi, regional dialects as core product (Industs AI, Quantis AI)
• Accent Support — Australian, British, American English (Industs AI)
• Foreign Languages — Arabic, German, French, Mandarin (in development) (Industs AI)

Video Analytics & Geospatial

• Petabyte-Scale Video Processing — 100+ PB of video data processed in real time (Papy Labs)
• Vehicle Dash Cams & CCTV — Data sources (Papy Labs)
• Object Detection & Counting — Passenger counting, goods counting, vehicle categorization (Papy Labs)
• Dynamic Layering via LLM — Semantic search over video (e.g., "find all instances where helmets are missing") (Papy Labs)
• Tile-Based Geospatial Pricing — 25km² tiles at ₹1.5L/day (Papy Labs)
• Real-Time Comparison — Track changes over time (6 days, 1 year) (Papy Labs)

Robotics & Industrial

• Industrial Robots — World's largest manufacturer deployment context (Technoate AI)
• CNC Programming — Automated program generation for machine tools (Technoate AI)
• Shop Floor Automation — 100% automation without human supervision (Technoate AI reference: ABB/similar)
• Robot Integration & Commissioning — End-to-end deployment automation (Technoate AI)

Infrastructure & Integration

• Cloud Hyperscalers — AWS, Google Cloud, Azure mentioned as benchmarks; many Indian startups avoiding for data residency (Papy Labs, Industs AI)
• GPUs & Compute — CAD (Compute as Data), Aravat mentioned as Indian sovereign compute providers (Papy Labs)
• WebSockets & API Integration — Call