Chebrolu Trilokesh
Venkata Uday

VoteEdge

AI-Powered Election Campaign Management Platform (Nigeria)

• Launched an 8-module, multi-tenant election-campaign platform covering 93.4M+ voters, 176,974 polling units and 774 LGAs in 4.5 months.
• Cut LLM cost with multi-LLM routing (Claude, GPT-4o, Gemini via LiteLLM) and Redis semantic caching, powering OCR result collation, voter scoring and sentiment analysis without custom model training.
• Built the Kafka → Spark + Delta Lake pipeline (Elasticsearch search, PostgreSQL/PostGIS geospatial) on a multi-region AWS deployment with DR.
• Shipped an offline-first React Native app (WatermelonDB, CRDTs) with 72-hour offline support for field agents.

GURU

AI Teaching Operating System

• Designed a 7-engine, ~19-agent AI teaching system coordinated through a central Student Intelligence Graph, for K-12 and competitive-exam learners.
• Built CARA, a voice-first Socratic tutor supporting 22+ Indian languages, and SOMA, a live student model tracking per-concept mastery, cognitive load and learning style.
• Built EVAL (confidence-weighted scoring, board-exam forecasting) and MEMO (SM-2/FSRS spaced repetition with auto-generated flashcards).

Model Verse

B2C AI Marketing Asset Generation Platform

• Cut marketing-asset production from weeks to minutes and cost by ~90% versus traditional photoshoots by generating assets directly from a brand URL.
• Designed the microservices backend (gateway plus User, Project, Brand, Asset, Billing) with async generation workers on PostgreSQL, Redis and S3-compatible storage.
• Integrated image, virtual try-on and video providers with fallback routing, plus GPT-4 Vision/Claude for brand and competitor analysis.

Self-Service Analytics Platform (Google BigQuery)

• Delivered a multi-tenant self-service analytics platform in 24 weeks, combining RAG document Q&A with NL-to-SQL over BigQuery via a React app and Slack bot.
• Kept latency low: RAG pipeline 1.2s avg, BigQuery layer 850ms avg, with pgvector and Pinecone retrieval.
• Enforced tenant isolation with RBAC and enterprise SSO (Okta, Azure AD, Google Workspace); added SQL transparency with confidence scores and source citations.

Snowflake Agentic Analytics Pipeline

• Built a LangGraph/CrewAI multi-agent pipeline from intent routing to SQL generation, validation, execution and insight/dashboard creation.
• Made NL-to-Snowflake-SQL self-correcting: queries are validated in LIMIT-100 mode and re-prompted on error, backed by a semantic layer of facts, dimensions and KPIs.
• Implemented governed writeback (validate → submit → approve → merge) on whitelisted views with row-access, data masking and full audit logging.