Skip to content

Open to roles in distributed systems and AI infrastructure

Shreyas Patange

Software engineer building distributed systems and agentic AI.

Download PDF

Experience

Real-time video infrastructure, agentic systems, and the platform work that keeps both running.

  1. Jul 2025 – Present

    1 yr 4 mo

    Skylark Labs

    Software Development Engineer IInow

    • Built a multi-region distributed system for RTSP/WebRTC video ingestion, egress, and processing in TypeScript, Go, OpenCV, FFmpeg, and gRPC, sustaining 10K+ concurrent live streams with real-time analytics at sub-400ms p95 end-to-end latency.
    • Developed an agentic AI platform with Mastra and TypeScript: hybrid retrieval (metadata filtering, semantic search, structured querying), tool calling, contextual memory, and autonomous workflows, enabling natural-language analytics over 1M+ surveillance events and cutting manual review time by 60%.
    • Implemented Envoy and Traefik load balancing for gRPC and HTTP services with Maglev consistent hashing, distributing 10K+ live streams across processing nodes and enabling A/B deployments on Kubernetes and Docker.
    • Designed a centralized identity and access management platform with OAuth2, JWT authentication, organization management, licensing, and fine-grained RBAC, serving multiple internal services and handling 100K+ auth and licensing requests monthly.
    • Managed multi-region AWS infrastructure (Kubernetes, ECS, Lambda, RDS) behind Cloudflare global load balancing, with Terraform IaC, CI/CD pipelines, and observability via Prometheus, Grafana, and Loki.
    • Improved reliability with AWS SQS/SNS distributed queues and built the real-time alerting stack (WebSockets, push, email, SMS), delivering 100K+ alerts/day at under 2s median latency.
  2. Dec 2023 – Jul 2025

    1 yr 8 mo

    Heliverse Technologies

    Tech Lead

    • Led a team of 5 engineers; built an AI-powered real-time voice agent and WhatsApp agent with LangChain and Twilio, handling 100+ conversations/day through BullMQ workflows and automating 70% of inquiries end to end.
    • Developed a serverless microservice architecture on AWS Lambda with DynamoDB, TypeScript, and CloudFront, deployed via the Serverless Framework.
    • Optimized frontend performance with Next.js SSG/SSR and server actions for type-safe RPC calls.
    • Architected an HR platform backend with NestJS, Prisma, and PostgreSQL: shift management, biometric attendance, and payroll generation for 100+ employees, with real-time notifications via Firebase and WebSockets.

Projects

Side projects that go deeper than a weekend: a language with its own interpreter, and a multiplayer game with real consistency problems.

The Gor landing page: the language name, its tagline, and links to the playground and the docs

A Turing-complete interpreted programming language written from scratch in Go.

Full tokenizer, parser, AST, optimizer, evaluation engine, and an interactive REPL. An online playground built with React and Gin lets you write and run Gor code and inspect the AST it produces.

  • Go
  • Gin
  • React
GeoQuiz round in progress with a Street View panorama and the guess map

Street View location-guessing game with real-time multiplayer lobbies.

Built with TypeScript, React, the Google Maps Street View API, and Supabase. Multiplayer runs over Supabase's WebSocket-driven realtime streams with synchronized room state, competitive rounds, chat, reconnection handling, and deterministic scoring in PostgreSQL so every client agrees on the result.

  • TypeScript
  • React
  • Redux
  • Supabase
  • PostgreSQL
  • WebSockets

Earlier projects

About

How I got here and what I reach for.

I'm a software engineer at Skylark Labs working on distributed systems and agentic AI. I built a multi-region video platform that handles thousands of concurrent live camera streams, then built an agentic layer on top of it that lets people query more than a million surveillance events in plain language.

On the infrastructure side I own the real-time pipeline: RTSP ingest, FFmpeg and OpenCV processing, gRPC service boundaries, WebRTC delivery, and the Envoy and Maglev routing that keeps streams pinned to nodes without hotspots. On the AI side I work in TypeScript with Mastra, combining hybrid retrieval, structured queries, tool calling, and memory.

Agents are only as reliable as the systems underneath them, which is why I like working on both. I do my best work on small teams with real ownership. Outside work I built Gor, an interpreted language in Go, and I compete on LeetCode.

Education

University of Mumbai

B.E. in Computer Science, minor in Data Science

Aug 2020 – May 2024GPA 9.30 / 10.00

Competitive programming
LeetCode Knight (top 5%) with a 2000+ contest rating and 1100+ problems solved. Profile
Based in
Mumbai, India

Skills

Brighter tags are the tools I reach for most weeks.

Languages

  • TypeScript
  • JavaScript
  • Go
  • Python
  • C/C++
  • Rust
  • Zig
  • Bash
  • SQL

Frameworks & technologies

  • React
  • Next.js
  • NestJS
  • FastAPI
  • Gin
  • Hono
  • Bun
  • GraphQL
  • gRPC
  • WebRTC
  • WebSocket
  • WebTransport
  • FFmpeg
  • NATS
  • Kafka
  • RabbitMQ
  • Temporal
  • LangChain
  • Mastra

Infrastructure & DevOps

  • Docker
  • Kubernetes
  • Envoy
  • Traefik
  • NGINX
  • Terraform
  • Helm
  • Linux
  • GitHub Actions

Databases

  • PostgreSQL
  • CockroachDB
  • MySQL (Vitess)
  • MongoDB
  • DynamoDB
  • ScyllaDB
  • Cassandra
  • Redis

Cloud

  • AWS
  • ECS
  • Lambda
  • RDS
  • S3
  • EC2
  • Cloudflare
  • GCP
  • Firebase
  • Supabase

Exposure

  • Java
  • C#
  • Spring Boot
  • .NET
  • Lua
  • WASM
  • ONNX
  • ONVIF
  • NVIDIA Triton

Concepts

  • Distributed systems
  • Consensus
  • System design
  • Load balancing
  • Microservices
  • Networking
  • Operating systems
  • Databases
  • Algorithms & data structures
  • RAG
  • Agentic AI

Say hi

Have a role, a project, or a hard systems problem? Send a message here or write to me directly.

patangeshreyas@gmail.com