Abhilash Bandari

Backend Developer · Go, Python, Node.js & AWS

I build low-latency trading and market-data systems — the kind where every millisecond and every dropped packet matters. Currently architecting broadcast and trading systems at SMC Global Securities; previously shipped high-volume logistics and finance backends at BYJU'S.

View my work Get in touch
Abhilash Bandari
Market-feed latency
40

ultra-low latency broadcast system built in Go & C#

Time-series records
8

on self-hosted TimescaleDB I architected, serving chart APIs at 30ms p99

Packet drops cut
65

via Linux kernel-level network tuning

Annual cost saving
₹2

by replacing a vendor product with the in-house charts platform

01

About

I'm a backend developer with 5+ years of experience designing and running high-performance systems in production — trading platforms at a financial services firm, IoT data pipelines, and edtech services handling 105,000+ daily shipments and $1M+ in monthly transactions.

My sweet spot is the layer where application code meets the network and the OS: UDP multicast feeds, gRPC and WebSocket services, kernel tuning, cloud network architecture on AWS (Direct Connect, Transit Gateway, GRE tunnels), and time-series data at scale — I designed and operate a self-hosted TimescaleDB holding 8B+ records behind 30ms-p99 chart APIs. I lead a team of four developers and write about what I learn on Medium.

Languages

  • Go
  • Python
  • C#
  • JavaScript / Node.js
  • Shell

Data & Search

  • PostgreSQL
  • TimescaleDB
  • MongoDB
  • MSSQL
  • Redis
  • Apache Solr
  • InfluxDB

Cloud & Infra

  • AWS (EC2, ECS, S3, Lambda, NLB/ALB, Transit Gateway, VPC, WAF)
  • Docker
  • Kubernetes
  • CI/CD

Protocols & Realtime

  • gRPC
  • WebSockets
  • UDP Multicast
  • REST
02

Experience

Aug 2023 — Present Delhi, India

Member of Technical Staff–3

SMC Global Securities · financial services / trading

  • Architected a high-performance broadcast / market-feed system in Go and C#, delivering real-time exchange data with ~40ms end-to-end latency.
  • Architected an in-house charts platform end to end on self-hosted TimescaleDB — schema, ingestion and query layer for 8B+ time-series records, serving chart APIs at 30ms p99 and saving ₹2 Cr per year in vendor costs.
  • Led the migration of on-premise trading systems to AWS using Direct Connect, GRE tunnels and Transit Gateway; resolved packet drops with Linux kernel tuning, cutting drops by 65%.
  • Developed and scaled a stateless GTT order system and price-alerts service in Go.
  • Tuned Apache Solr search to a p99 of 70ms; rebuilt the BOD batch process in Python, cutting runtime from 4 hours to 30 minutes.
  • Provide technical leadership to a team of 4 developers.
Jun 2023 — Aug 2023 Remote · Vadodara

Software Engineer

KorelOT Solutions · IoT

  • Built a microservice migrating stored documents to Azure Blob Storage, trimming monthly storage costs.
  • Optimized client-facing APIs, dropping p99 latency from 960ms to 98ms.
Apr 2021 — Jun 2023 Remote · Bangalore

Software Engineer

BYJU'S (Think & Learn) · edtech

  • Integrated logistics APIs processing 105,000+ daily shipments; built the job-orders system in Go.
  • Built microservices for Supply Chain and Digital Finance verticals handling $1M+ in monthly transactions.
  • Shipped real-time analytics dashboards (React + Node.js) and a customer-facing book-delivery tracking platform; led a team of 4 developers.
03

Open-source projects

Go libraries for exchange market data, plus tools for networking, infrastructure and the everyday development workflow.

Pure Go Compression library

lzo1z / LZO-GO

LZO1Z decompression for exchange feeds and legacy protocols. Decodes into caller-owned buffers with explicit errors for malformed input. No cgo or dependencies outside the standard library.

  • 0 allocations per call in the published benchmarks
  • Validated against vectors generated by the original C library
Source ↗ API docs ↗ Story & benchmarks →
GPL-2.0-or-later
Pure Go Market data library

fastfix

FAST 1.1 decoding with runtime XML templates, groups and sequences. Preserves decimal prices and gives feed handlers control over dictionary resets. No cgo or dependencies outside the standard library.

  • All six FAST field operators supported for decoding
  • Benchmarked against Go, C++, Java and Rust implementations
Source ↗ API docs ↗ Story & benchmarks →
MIT

Tools & developer setup

Go

connection-dump

A friendlier alternative to tcpdump: monitor TCP and UDP traffic by port on any interface, with configurable reporting intervals. Cross-platform, single binary.

Read the write-up →
Go

multicast-test-tool

CLI for verifying UDP multicast connectivity between a source and multiple destinations — list interfaces, send and receive test messages. Invaluable when debugging trading networks.

Read the write-up →
Shell · macOS

dotfiles

My macOS setup for keyboard-driven window tiling with AeroSpace, workspace and app indicators in SketchyBar, and switchable themes. Includes shortcuts for moving windows between workspaces and monitors, with configs managed through GNU Stow.

Explore the setup →
04

Writing

Notes from production — networking, search and infrastructure.

  • macos
  • productivity
  • dotfiles

Every App Has an Address: A Keyboard-Driven Mac for Production Incidents

AeroSpace and SketchyBar give every app a fixed place one keystroke away, so a shared screen on an incident call never turns into a window hunt.

Read →
  • go
  • market-data
  • open-source

Two Formats, One Language: Decoding LZO and FAST Market Data in Pure Go

How FAST decoding and LZO decompression moved our feed processing from about 32 CPU cores to under one, with LZO1Z benchmarks against C and Java.

Read →
  • go
  • monitoring
  • tcpdump

Network Traffic Monitoring: From Zero to Hero in 10 Minutes

Introducing nwdump, a Go-based traffic monitoring tool built as a modern, cross-platform alternative to tcpdump for TCP/UDP monitoring in production.

Read on Medium →
  • apache-solr
  • optimization
  • redis

The Hidden Landmines: Configuration Pitfalls That Will Destroy Your Solr Production Setup

Critical Solr configuration mistakes — managed vs. classic schema confusion, improper commit settings — and how to avoid them in production.

Read on Medium →
  • docker
  • aws
  • multicast

Docker and Multicast in AWS: A Deep Dive into Network Broadcasting

Multicast networking inside Docker containers on AWS — host network mode, IGMP configuration and Transit Gateway multicast architecture.

Read on Medium →
  • golang
  • udp
  • network

Multicast Testing Tool

An open-source Go tool for testing multicast communication across networks — interface listing, message send/receive and cross-platform binaries.

Read on Medium →
All posts on Medium
05

Get in touch

Happy to exchange ideas about low-latency systems, Go, infrastructure and open-source projects.

bandariabhilash0@gmail.com
GitHub · LinkedIn · Medium