K. Vedant MahajanLead Software Engineer | Node.js · Java · PostgreSQL · Applied AI
Profile

Lead Backend Engineer with 10+ years of experience designing and delivering production systems at scale. At Khatabook, built event-driven pipelines processing 8.5 million daily SQS transactions, rule-based allocation engines, and geospatial services in Node.js, Java 17, and PostgreSQL. At Toppr/Byju's, led a team of 3–5 engineers as Principal Engineer scaling India's #1-ranked Q&A product to 35M MAU — outperforming Byju's own product and contributing to the acquisition. Brings hands-on Applied AI and Generative AI engineering practice — RAG pipelines, semantic embeddings, pgvector similarity search, LangGraph agents, and LLM API integration built in Node.js/NestJS.

Awards

AIR 119, Global 1709.

Feb 2020
Skills
Backend Ecosystem — Node.js, NestJS, Express.js, PostgreSQL, PostGIS, Redis, SQL, Kafka, Amazon SQS, Java, Spring Boot, Spring Data JPA, Resilience4j, AWS S3, Kubernetes
Languages — JavaScript/Typescript, Java, Python, C++, PHP
Frontend Ecosystem — React.js, Next.js, Redux, React Query, HTML/CSS, SCSS, TailwindCSS, Webpack
Observability — Coralogix, Grafana, Lens, AWS cloudwatch, Datadog, Sentry, New Relic
System Design & CS Fundamentals — High-Level Design (HLD), Low-Level Design (LLD), Data Structures & Algorithms, Microservices Architecture, Event-Driven Architecture, API Design Patterns, Database Query optimisations, Caching Strategies, Message Queue Patterns (SQS, Kafka), WebSockets, Web Security (OWASP), Web Performance Optimisation
Applied AI & Generative AI Engineering: — RAG Pipeline Design, Semantic Embeddings, Vector Similarity Search (pgvector), LLM API Integration (OpenAI, Anthropic), Prompt Engineering, LangChain, Agentic AI Workflows, Generative AI Application Architecture, AI-Assisted Development (Cursor)
Leadership & Collaboration — Technical Mentorship, Team Leadership (3–5 engineers), Code Review, Technical Roadmapping, Architecture Decision Records (ADRs), Cross-functional Collaboration, Agile/Scrum Delivery
Professional Experience

SDE 3 (Backend)

Khatabook⁠
  • •Architected and delivered an automated IVR calling worker (NestJS) over a multi-tenant SQS-based notification pipeline processing 8.5M daily messages — implementing per-tenant Exotel rate limiting (Redis token bucket), transactional bulk assignment, and async webhook-driven status tracking across COMPLETED / NO_ANSWER / FAILED states — directly replacing a manual CS calling operation and enabling the company to downsize and relocate the CS function from Bangalore to Jaipur, significantly reducing operational overhead
    • Apr 2025 – May 2026Bangalore, India
  • •Engineered a dual-mode search system for the sales lead platform — implemented full-text search via pg_trgm (GIN-indexed trigram matching) for static field lookup across lead name, business name, and phone number, and spatial search via PostGIS with configurable distance filters (0–50km default radius, custom range support), reducing query latency from 500ms to 20–50ms and enabling field reps to surface relevant leads 10–25x faster.
  • •Contributed to a Java 17 / Spring Boot digital lending microservice (Khatabook Loans) — extended the configurable loan journey engine by implementing a new step processor following the StepProcessorInterface contract (preprocess → validate → update lifecycle). Worked across a codebase handling full loan lifecycle: origination, multi-NBFC lender routing, NACH repayments, and settlement reconciliation. Stack: Java 17, Spring Boot, Spring Data JPA, PostgreSQL, Redis, Resilience4j, AWS SQS/S3
  • •Initiated exploration of semantic search and RAG-based retrieval patterns as a potential enhancement to the lead discovery layer — prototyped in parallel as an independent engineering practice (see Projects section)

    • SDE-2 ( Full stack )

    Khatabook
  • •Enhanced multi-constraint lead allocation engine: Steered a rule-based allocation system balancing language matching, workload distribution, and sticky allocation. Refined dynamic SQL query generation and allocation configs, overcoming single-worker architecture limitations to ensure strong consistency
    • Dec 2023 – Mar 2025Bangalore, India
  • •Streamlined Concurrent Bulk Operations: Executed bulk assignment processing with controlled concurrency and per-assignment transaction isolation, with comprehensive error handling, audit logging, and CSV-based tracking
  • •Built Underwriting (UW) Task Flow from scratch: Designed and implemented a rule-based task creation system and engineered a task allocation worker with priority-based sorting. Codified 7+ business rules using a rules engine pattern where rules execute independently, and architected a multi-step verification workflow using a Strategy pattern for task types (UW, Fraud) to enable extensible workflows.
  • •Optimized lead management system serving 2 client applications (mobile app, web dashboard) with integration to upstream loan origination and downstream notification services, processing thousands of leads daily

    • FullStack Engineer (Platform Team)

    Quillbot⁠

    Contributed to the end-to-end feature delivery for an enterprise-wide CMS product. Improved system stability and application reliability by driving a complete architectural overhaul of frontend state management flows

    Apr 2023 – Aug 2023Remote

    Principal Engineer

    Byjus⁠
  • •Led engineering of Toppr's Q&A product to India's #1 educational app ranking (SimilarWeb, 35M MAU) — outperforming Byju's own product, directly contributing to Toppr's acquisition. Owned full technical roadmap across performance, stability, and feature delivery for a team of 3–5 engineers.
    • Jul 2022 – Mar 2023Remote, India
  • •Architected AskAnExpert as modular npm packages in a monorepo — converting a monolithic Q&A product into a plug-and-play integration layer adopted across Toppr Ask, Byju's QnA, and internal products, driving a high-revenue monetisation stream.

    • Tech Lead

    Byjus
  • •Led a team of 3–4 engineers resolving critical memory overflow issues causing 5xx errors on a Next.js platform serving lakhs of content pages across text, media, and video formats. Drove continuous page speed and bundle size optimisation pre-deployment via load testing (npm loadtest) and pm2 stage monitoring.
    • Sep 2020 – Jun 2022
  • •Built CI/CD pipelines with automated linting gates, failing builds on threshold breaches, and Bitbucket webhook integrations with Slack and email for pre-deployment quality enforcement. Used Screaming Frog crawler for post-production page crawl health monitoring.
  • •Mentored 3–4 junior engineers including participation in performance review cycles.

    • Senior Software Engineer

    Turtlemint⁠
  • •Built the foundational frontend infrastructure at Turtlemint — server-driven configurable checkout flows, the core component library @turtlemint/mint-ui ( a monorepo), and @turtlemint/redux-analytics for automating event tracking through Redux actions — enabling consistent, scalable development across vertical teams.
    • Jul 2019 – May 2020Mumbai, India

    Software Engineer

    Turtlemint
  • •Contributed to core platform optimisation as part of the platform team — iteratively built a React.js boilerplate benchmarked against the popular react-boilerplate open source standard, achieving superior performance and smaller bundle size through measured, incremental library and tooling additions
    • Dec 2018 – Jun 2019Mumbai, India

    Software Engineer | Freelance → Alfa KPO → Justdial → NSE IT → Cogoport

    Progressed through engineering roles across freelance web development, React, Angular, Ionic, and Node.js.

    Dec 2013 – Jul 2018Mumbai, India

    Built client websites, email marketing portals, hybrid mobile apps (Firebase push notifications), admin

    dashboards, and web portals.

    Education

    B.Tech - Major in Computer science engineering

    I.K. Gujral Punjab Technical University⁠
    2009 – 2013Gurdaspur, India
    Projects

    AI Engineering Projects | Independent (2026–present)

  • •Production RAG Pipeline — Built end-to-end Retrieval Augmented Generation system in Node.js/NestJS: semantic chunking, pgvector similarity search (cosine), augmented prompt construction with XML structure, streaming SSE endpoint with citation parsing, Zod schema validation with retry logic, and confidence threshold gating. Prevents hallucination by constraining LLM to retrieved context only.
    • Jan 2026 – Present
  • •Multi-tool LangGraph Agent — Built a stateful ReAct agent using LangGraph.js with multiple tool definitions, PostgreSQL checkpointing for state persistence, human-in-the-loop approval node for irreversible actions, and conditional routing between tool execution and final answer generation.
  • •Semantic Search Service — NestJS service using all-MiniLM-L6-v2 embeddings and pgvector. Model loaded once at onModuleInit to prevent thundering herd on restart. SQL-level threshold filtering to use pgvector index correctly. Returns similarity confidence score alongside results.

    • Independent Project — Real Estate Tech Platform | Sep 2023 | Bangalore

    Built a real estate discovery platform integrating RERA data pipelines for automated fact-checking, a multi-signal Trust Score algorithm, and an AI-powered preference matching layer using semantic embeddings and vector similarity search. Applied RAG-style retrieval architecture hands-on