About me
I'm a software engineer building backend systems, AI infrastructure, and developer tooling — with a focus on the architecture of AI-integrated backends and scalable platforms: workflow orchestration, retrieval, execution infrastructure, and the engineering that keeps production AI systems reliable.
Most of my work involves building production systems, but I'm broadly interested in how software behaves under real-world constraints — what changes under concurrency, where state lives through execution, how systems recover from failures, where bottlenecks emerge, and how architecture evolves as complexity grows. I prefer systems that stay reliable, observable, and easy to reason about.
When something breaks, I trace it to the root cause through logs, dependencies, and failure paths, and I lean toward making structural decisions early — around component interactions, failure containment, and scalability — over clever fixes that get hard to reason about later. Today that mostly means multi-agent workflows and running AI systems in production: context engineering, execution efficiency, and the guardrails needed to operate them safely at scale.
Beyond work, I've contributed to open-source projects including Langflow and AiiDA, and I keep coming back to Living Dead — a long-term game project I'm gradually expanding as I learn more about gameplay systems, reinforcement learning, multiplayer networking, and engine architecture.
What i'm doing
Web Engineer
Game Development
'%3e%3cpath%20d='M5,2c-1.64453,0%20-3,1.35547%20-3,3v30c0,1.64453%201.35547,3%203,3h14v2h-9c-0.03125,0%20-0.0625,0%20-0.09375,0c-0.26953,0.02734%20-0.51953,0.16016%20-0.6875,0.375l-5,6c-0.24219,0.30078%20-0.28906,0.71094%20-0.12109,1.05859c0.16797,0.34766%200.51953,0.56641%200.90234,0.56641h40c0.38281,0%200.73438,-0.21875%200.90234,-0.56641c0.16797,-0.34766%200.12109,-0.75781%20-0.12109,-1.05859l-5,-6c-0.19141,-0.23828%20-0.47656,-0.375%20-0.78125,-0.375h-9v-2h14c1.64453,0%203,-1.35547%203,-3v-30c0,-1.64453%20-1.35547,-3%20-3,-3zM5,4h40c0.55469,0%201,0.44531%201,1v30c0,0.55469%20-0.44531,1%20-1,1h-40c-0.55469,0%20-1,-0.44531%20-1,-1v-30c0,-0.55469%200.44531,-1%201,-1zM6,6v28h38v-28zM8,8h34v24h-34zM19.28125,11.28125l-8,8l-0.6875,0.71875l0.6875,0.71875l8,8l1.4375,-1.4375l-7.28125,-7.28125l7.28125,-7.28125zM30.71875,11.28125l-1.4375,1.4375l7.28125,7.28125l-7.28125,7.28125l1.4375,1.4375l8,-8l0.6875,-0.71875l-0.6875,-0.71875zM21,38h8v4h-8zM10.46875,42h8.53125v2h12v-2h8.53125l3.34375,4h-35.75z'%3e%3c/path%3e%3c/g%3e%3c/g%3e%3c/svg%3e){kind=small}
Backend Engineering
Agentic AI Systems
Machine Learning
Open-Source Contributor
Experience
Evalueserve
AI/ML Engineer | Platform Engineering - Python
- FastAPI
- LangChain
- MCP
- LLMOps
- Multi-Agent Systems
- PostgreSQL
- pgvector
- ChromaDB
- Redis
- Azure
Platform: Genie — an enterprise Agentic AI Studio for designing, executing, and monitoring complex AI workflows.
- Reduced workflow execution latency by ~60% on benchmark workflows by engineering a Python/FastAPI DAG orchestration engine with concurrent node scheduling, asynchronous execution, fault-tolerant routing, shared state, Pydantic interfaces, and Redis-backed checkpointing.
- Architected a modular GenAI component ecosystem powering multi-agent orchestration, RAG, OCR parsing, metadata-aware chunking, memory systems, MCP integrations, LLM-as-a-Judge, and human-in-the-loop workflows via a ReactFlow editor with SSE streaming.
- Designed session-scoped context infrastructure using PostgreSQL (pgvector, JSONB) and ChromaDB, with hybrid retrieval, cross-encoder re-ranking, source deduplication, embedding reuse, and semantic caching to improve retrieval precision and token efficiency.
- Scaled platform execution capacity by 2× while reducing process memory footprint (2.2GB to 755MB) and cold-start latency (~30s to ~8s) by introducing process-isolated Celery worker execution and lazy component loading.
- Hardened platform governance and reliability by implementing authenticated APIs, LOB-scoped token quotas, security guardrails, execution gating, automated tracing and observability, backed by 93% pytest coverage.
Arena
Software Engineer (Consultant) - React
- TypeScript
- NestJS
- REST APIs
- PostgreSQL
- Redis
- React Query
- TypeORM
Product: SocialFi Platform — a Web3 platform for social engagement, creator monetization, and blockchain interactions on Avalanche.
- Built full-stack features and type-safe REST APIs (React, TypeScript, NestJS, PostgreSQL) for social feeds, real-time messaging, and streaming.
- Optimized high-throughput backend performance with Redis caching, efficient TypeORM queries, selective response shaping, lazy-loading pagination, and atomic counters for consistency under concurrent writes.
- Improved frontend responsiveness and stability with custom React Query hooks, optimistic updates, and targeted cache invalidation, resolving synchronization issues across PostgreSQL and React rendering.
Publicis Sapient
Software Engineer Internship - Node.js
- Express.js
- MongoDB
- Mongoose
- RabbitMQ
- JWT
- Pug
- REST APIs
- Microservices Architecture
Project: Hospital Management Platform — a distributed, event-driven architecture built from independently deployable microservices.
- Architected a Node.js/Express Patient History microservice with JWT authentication and Mongoose, using Pug server-side rendering to eliminate client-side fetch cycles and reduce exposed API surface.
- Integrated RabbitMQ inter-service asynchronous communication with acknowledgments and idempotent, fault-tolerant event processing.
