Tech Lead | Blockchain, AI, Cloud Platforms
Bengaluru, India
lakshyakumar.github.io/lakshyakumar
Engineering leader with 7+ years across fintech, AI, and Web3. I've shipped a Polygon and Ethereum payments platform doing 25,000+ transactions a day, an institutional fund tokenization platform on EVM and Provenance, and an enterprise RAG assistant that's now in production. These days I'm tech lead at Payram, building ZepaPay, an on-ramp/off-ramp crypto payments platform, where I also own the AI agents behind support and operations. I move between architecture review and Go concurrency code without much friction, and I've grown engineering teams from five to fifteen while staying hands-on. Core maintainer of Hyperledger Bevel, author of Mjolnir, and I write on Medium.
- Current role: Tech Lead (Contract) at Payram, a self-hosted crypto payments platform in Go (Nov 2025 – Present)
- I own the architecture, implementation, and production rollout of the payments platform.
- Built and shipped a live wallet and payments app handling real transactions.
- Tech lead on ZepaPay, an on-ramp/off-ramp crypto payments platform built for Consult Now on Payram's rails. It handles fiat-to-crypto conversion, KYC and compliance, and settlement. I also built and run the AI agents behind its support and operations flows.
- Previous role: VP, Web3 & AI Solutions at Formidium (Feb 2024 – Oct 2025)
- Ran containerized deploys through AWS CodePipeline (CodeCommit, GitHub Actions, ECR, ECS, CodeBuild) with infrastructure defined in Terraform and CloudFormation.
- Architected a fund tokenization platform on EVM and Provenance, and built the enterprise RAG assistant (LangChain, LangGraph, Next.js). Retrieval mixed keyword and vector search with a reranking pass, and I put the whole thing behind an eval suite so quality regressions got caught in CI before they shipped.
- Education: M.Tech in Computational Mathematics, NIT Surathkal
- Core strengths: blockchain architecture, AI/RAG systems, cloud infrastructure, production reliability, team leadership
- Deep interests: crypto rails, ERC-4337 account abstraction, key custody, ZKPs, MCP tooling, developer productivity
Languages JavaScript, TypeScript, Python, Go, Solidity, Rust
Backend & Frameworks Node.js, React, Next.js, REST APIs, Microservices, LangChain, LangGraph, CrewAI, PydanticAI
Blockchain EVM, Provenance Blockchain, Hyperledger Fabric, Hyperledger Besu, Polygon, Avalanche, ERC-1400, ERC-4337, Account Abstraction, Smart Contract Wallets, Tokenization, Cross-chain Bridges
AI & Agent Systems RAG, Hybrid Retrieval (BM25 + dense, RRF), Reranking, Citations/Faithfulness, Semantic Search, Vector Databases, AI Memory Systems, MCP, MCP Servers, AI Tool Orchestration, Agent Guardrails, Prompt-Injection Defense, Model Routing, Context Management
Evals, Fine-tuning & LLMOps LLM-as-Judge, Trajectory Evals, Golden Datasets, CI Eval Gates, Drift Detection, Fine-tuning (LoRA / QLoRA, PEFT), Model Serving (vLLM, Triton, TensorRT-LLM), Quantization (AWQ/GPTQ/FP8), MLflow, Langfuse, p95/p99 SLOs
Cloud, DevOps & Reliability AWS (ECS, EC2, Lambda, S3, ECR, CodeBuild, CodePipeline), Docker, Kubernetes, GPU Orchestration, Autoscaling, GitHub Actions, Jenkins, CI/CD, GitOps, CI Eval Gates, Zero-downtime deployments, Failover design, Monitoring, Alerting
Leadership & Delivery Technical Architecture, Team Leadership, Mentorship, Cross-functional Collaboration, Delivery Management
Hands-on learning platform for builders, with 7,500+ challenges across 14 tracks covering AI/ML, blockchain, cloud, security, and systems. It plugs into AI assistants: Claude, Cursor, or any MCP-aware agent can pull the course material over a read-only API to teach and quiz you, in 40+ languages.
Repository Intelligence MCP server for coding agents. Provides structured codebase context, semantic search across repos, and tool-based access patterns for AI-assisted development workflows.
Full-spectrum market intelligence Claude Skill for founders — Lean Canvas extended with competitor analysis, validation frameworks, and strategic clarity. Accepted use in the Claude skills ecosystem.
Context-aware AI agent built on LangGraph and PydanticAI. Searches the web, answers queries, builds knowledge bases, and maintains conversational context with token budgeting and dynamic subgraph switching.
FastAPI-based AI agent that scrapes URIs, answers queries, and creates persistent knowledge bases using LangGraph and PydanticAI.
Production AI assistant for document question answering using LangChain and LangGraph. AI rephrases and simplifies queries before vectorized search for improved retrieval accuracy.
Curated learning resource and reference implementations for LangGraph-based agent systems.
Live web application for structured technical challenges and tooling experiments.
MCP server implementation in TypeScript for structured AI tool delivery and context management.
AI agent for movie discovery, recommendations, and knowledge retrieval workflows.
On-ramp/off-ramp crypto payments platform built on Payram's rails for Consult Now. Handles fiat-to-crypto conversion, KYC and compliance, and settlement. I was tech lead, and also built the AI agents that run behind support and operations.
Open-source accelerator for production-ready distributed ledger deployment on Kubernetes. Contributed deployment pipelines for Fabric, Besu, and Quorum.
Fund tokenization platform across EVM-compatible and Provenance chains using ERC-1400 — staking, sales, and token contract architecture.
Step-by-step exploration and practical implementation of smart contract wallets and account abstraction patterns using Solidity.
Lightweight TypeScript/ethers.js wrapper for cleaner smart contract interaction and improved developer ergonomics. Published to npm.
React and Next.js hook library for blockchain interfaces and wallet-connected applications. Published to npm.
Production-ready Next.js + Tailwind + Mongoose starter with authentication, TypeScript, and clean project structure.
Enterprise-grade NestJS boilerplate with Mongoose, TypeScript, and modular architecture.
Comprehensive interview preparation roadmap covering system design, data structures, and engineering concepts. Live on GitHub Pages.
I write about hard technical ideas in ways engineering teams can actually use. Topics include:
- Crypto payments infrastructure and settlement workflows
- ERC-4337 account abstraction and smart contract wallet design
- Blockchain architecture, tokenization, and custody systems
- RAG systems, AI tooling, MCP servers, and developer automation
- AWS Certified Cloud Practitioner (Amazon Web Services)
- Building AI-Powered Search with MongoDB Vector Search (MongoDB, 2025)
- Building RAG Apps Using MongoDB (MongoDB, 2025)
- M.Tech in Computational Mathematics — NIT Surathkal
- B.Tech — Delhi Technological University
- Author of technical articles on account abstraction, AI tooling, vector search, blockchain systems, and developer tools.
- Active contributor to open-source blockchain and developer tooling ecosystems.
- Senior engineering, tech lead, solutions architect, and architecture roles in fintech, crypto, AI, and backend systems
- Advisory and product engineering work in payments, tokenization, and applied AI
- Conversations around crypto rails, developer tooling, custody infrastructure, and MCP/AI agent systems
- Portfolio: lakshyakumar.github.io/lakshyakumar
- LinkedIn: https://linkedin.com/in/lakshya-kumar-52735737
- Email: lklsquare@gmail.com
- GitHub: https://github.com/lakshyakumar