Apr 2024 – Mar 2026 · 1 yr 11 mos

• Vymo AI Agents Platform (Agent IQ)
• Built an agentic AI platform for autonomous voice/chat interactions - think vapi.ai but integrated into Vymo's business workflows.
• Three layers:
• Agentic AI layer: LLM-agnostic, supports both speech-to-speech (OpenAI Realtime, Gemini Live) and STT→LLM→TTS pipelines. Live transcription, custom guardrails, call outcome extraction (intent, sentiment, ability to pay, <other configurable>), configurable voice/pace, configurable telephony provider(exotel, twilio). Multi-MCP server integration for real-time business actions mid-conversation. Observability, monitoring builtin.
• Orchestration layer: Temporal-based durable collection workflows coordinating millions of loans - follow-up scheduling, business hours, retry logic, backpressure management, channel throttling. Solved thundering herd (all workflows waking at 9 AM) elegantly.
• MCP integration layer: Thin gateway exposing Vymo actions as LLM-consumable tools with auth/RBAC enforcement.
• Example flow: Borrower says "call me in 2 weeks" → agent checks repayment deadline via MCP (2 days) → negotiates → borrower agrees(shares payment link over sms) → system extracts PTP, schedules follow-up within business hours, updates loan status. Cancels followup if borrower pays. All autonomous.
• Vymo Data Platform & Report Builder
• Rebuilt Vymo's fragmented data infrastructure (MongoDB, ES, MySQL, Vertica, S3) into a unified ClickHouse-based analytics layer with multi-tenant isolation. Replaced flaky ETL pipelines with Debezium CDC - stable, observable, with point-in-time replay.
• Added Superset with RBAC for self-serve reporting. Introduced tiered storage for cost optimization.
• Results:
• Report shipping from 2-3 weeks → <1 day
• Deprecated Vertica: 50% cost savings, 3x query perf
• Data inconsistencies → <0.001%

Apr 2023 – Mar 2024 · 11 mos

• Vymo AI Copilot
• Spearheaded development of a conversational AI system with multiple applications:
• Built RAG pipelines using Milvus, Microsoft Guidance, Chat GPT 3.5 API and MongoDB
• Implemented domain-driven, contextual AI for BFSI sector
• Developed four specialized copilots:
• Training: Generated micro-courses from organizational content
• Analytics: Natural language interface for analytical queries
• Guided Selling: Intelligent interaction summarization and action integration
• Support: First-line defense for user queries, reducing ticket TAT from 45 to <2 minutes
• Developer Partner Platform (DPP):
• Led the design and initial development of DPP, a game-changing platform for Vymo:
• Enabled scalability of custom business workflows and improved external partner developer experience. We drastically reduced the time to market, de-cluttered the code base, reduced the cost of building custom business workflows and improved the stability of code
• Enabled external developer partners to build on top of Vymo stack with a good developer experience
• Utilized Temporal as workflow engine and Coder as remote development IDE
• Designed multi-tenant architecture supporting multiple programming languages
• Implemented support for version control, CI/CD, and RBAC of code written by external developers on a collaborative platform

Jan 2021 – Mar 2023 · 2 yrs 2 mos

• Built Vymo's ML team and platform from the ground up:
• Developed automated ML pipeline for data preprocessing, feature engineering, and result tracking
• Deployed 7 ML models to production, including:
• Lead Scoring: Custom models for each client, achieving 3x improvement in TTFM and 4x increase in policy issuance for hot leads
• ML-based Lead Allocation: Resulted in 8% increase in interested leads and 12.09% increased Lead Conversion Rate
• Next Best Actions (NBAs): Real-time recommendations with 30% acceptance rate
• Product Recommendation Engine: Context-aware suggestions for sales reps
• Agent Attrition Model: Predicted potential resignations 1-2 months in advance
• Partner Tiering and Inactivity Prediction: Advanced clustering & Anomaly detection for partner management

Apr 2019 – Dec 2020 · 1 yr 8 mos

• Goals Service:
• Designed, developed, and deployed service managing targets for 400k+ agents
• Built on Java + Spring Boot, scaled to handle 10k RPM on 2 nodes.
• Implemented self-serve UI for KPI creation and goal assignment
• Added support for external achievements, team goals, and composite achievements
• ETL Pipelines:
• Designed and implemented data sync from MongoDB to Vertica for user object data

Sep 2017 – Mar 2019 · 1 yr 6 mos

• Search Feature:
• Single-handedly built end-to-end full-text search capability
• Developed ETL pipeline from MongoDB to Elasticsearch
• Implemented APIs and UI for web and mobile (React Native) platforms
• Utilized advanced Elasticsearch concepts: dynamic templates, n-gram tokenizers, multi-language analyzers
• Supports 400k+ daily active users, stable in production for 6+ years
• Spearheaded development of key applications and services:
• iOS App:
• Led integration with Intunes MDM
• Self-serve UI Platform:
• Built foundation for form builder capturing data in Vymo app
• Developed UI framework using React and Redux
• Created auto-generated UI based on self-serve config type which facilitates CRUD operations on all self serve configs