Skip to content
AboutWorkSystemsContactDownload Résumé
Analyst Software Engineer · Zanskar Research

Souvik
Mukherjee.

I build the order management systems behind live Indian markets — order lifecycle, execution algos, and real-time risk, engineered for low latency.

GoC++20Low-LatencyLock-freegRPCTrading Systems
Get in touchRésumé
SENTINEL · SIMEX
LIVE FEED
PriceSizeTotal
100.402401,135
100.35168895
100.3096727
100.25132631
100.2084499
100.15205415
100.1060210
100.05150150
100.00▲ 0.00 (+0.00%)
spread 0.05
100.00142142
99.9596238
99.90212450
99.8564514
99.80176690
99.75120810
99.7088898
99.652381,136

simulated depth — the kind of feed Sentinel OMS & my matching engine drive

0+
Execution algos
bracket · OCO · trailing · iceberg
0×
Latency reduction
1000 ns → 100 ns per order
0+
Pre-trade checks
pluggable RMS compliance chain
0
Industry credentials
3× NISM · AWS SA · IEEE
01 / about

Engineering the order, end to end.

Analyst at Zanskar Research, where I lead development of Sentinel OMS — the firm's order management system. My work spans the full order lifecycle: state-machine-driven order flows, parent–child workflows, execution algorithms, real-time risk (RMS) validation, and mark-to-market accounting for low-latency trading. SEBI NISM-certified in Equity, Common, and Commodity Derivatives.

I care about systems that stay correct under pressure — where a missed response or a stale risk check has real consequences. That bias toward determinism, observability, and graceful recovery shows up in everything I build, from production trading infrastructure to side projects in C++.

01

Order lifecycle

State-machine-driven flows, parent–child workflows, and execution algos — from gRPC ingestion to exchange acknowledgement.

02

Real-time risk

Pluggable pre-trade compliance over deep-cloned portfolio snapshots, MTM accounting, and crash-safe recovery.

03

Low latency

Branchless hot paths, lock-free queues, cache-line layouts, and NUMA-aware pinning down to the sub-100 ns regime.

02 / work

Experience

Building and leading order management infrastructure at a proprietary trading firm.

  1. Analyst Software Engineer

    Zanskar Research · Bangalore, India

    May 2025 — Present
    • Led end-to-end design and development of Sentinel OMS v2 in Go for a live securities brokerage — defined architecture, owned technical decisions, and mentored 2 interns on order-lifecycle internals.
    • Architected a two-layer OMS separating user-facing parent orders from exchange-facing child orders; owned the lifecycle from gRPC ingestion through COLO bidirectional streaming, async NATS JetStream persistence, and real-time in-memory portfolio updates.
    • Built the OMS-side algo engine: bracket orders, OCO multi-exit monitoring, trailing stops with live LTP ratcheting, conditional entries, lot-size-aware iceberg (LCM) slicing, AMO/GTD scheduling, and freeze-chunk splitting for exchange quantity limits.
    • Designed a dual request-ID / response-ID replay protocol between OMS and COLO — on restart each side independently replays from its last-seen ID, guaranteeing exactly-once delivery and zero missed responses across pod restarts.
    • Built a pre-trade compliance chain of 15+ pluggable checks (margin, MTM, banned/RBI-restricted, gross qty/turnover, EDIS, rate limiting) over deep-cloned portfolio snapshots; instrumented µs-precision Prometheus latency tracking and crash recovery.

  2. Backend Intern

    Zanskar Research · Bangalore, India

    Dec 2024 — May 2025
    • Key contributor on the OMS team; developed limit, market, IOC, and iceberg order types with full lifecycle handling (new, modify, cancel) and exchange response integration.
    • Lead developer of the RMS (Risk Management System) module — enforced bans, restrictions, and MTM checks for robust risk controls.
    • Automated SPAN, VAR, and Refdata uploads to AWS S3 using Apache Airflow with calendar-aware scheduling based on NSE/BSE holidays.

Vellore Institute of Technology

B.Tech, Computer Science & Engineering

2021 — 2025

CGPA 8.9 / 10 · Tamil Nadu, India

03 / systems

Systems & projects

Production trading infrastructure and low-latency engineering, from microsecond matching to multi-agent reasoning.

FlagshipProprietary

Sentinel OMS v2

Production · Zanskar Research

A two-layer order management system in Go powering a live securities brokerage across Indian markets.

GogRPCNATS JetStreamPrometheusState Machines
  • gRPC ingestion → COLO bidirectional streaming
  • Async NATS JetStream persistence + in-memory portfolio
  • Algo engine: bracket / OCO / trailing / iceberg / AMO-GTD
  • Exactly-once replay protocol on pod restart

Order Matching Engine

C++20 · Low-Latency

An exchange-grade matching engine that pairs orders and streams tick-by-tick market data in microseconds.

  • Custom index-based Red-Black Tree, strict price–time priority
  • Branchless inlined fast-paths — no vtable stalls in <100 ns loops
  • Multicast UDP order flow + epoll non-blocking ingestion
C++20RB-TreeMulticast UDP
View source

AI Trade Journaling

Codeathon — 1st Prize (₹1L)

Hybrid deterministic + agentic system pairing FIFO-matched execution legs with OHLCV context for explainable per-trade insight.

  • CrewAI multi-agent reasoning, Pydantic-validated
  • FIFO leg matching + session-level analytics
  • MongoDB-cached, explainable trade narratives
CrewAIPythonMongoDB
View source

NetProbe

C++ · Networking

A low-latency TCP/UDP socket abstraction layer with non-blocking I/O and multicast membership control.

  • epoll event loop, connection pooling, adaptive batching
  • TCP_NODELAY, SO_REUSEPORT, recvmsg/sendmsg tuning
  • Unicast + broadcast transport, optional timestamping
C++epollTCP/UDP
View source

CoreBlur

C++ · Parallel Compute

A high-performance Gaussian blur engine using IPC and shared memory for zero-copy multi-process image processing.

  • Tile-based memory layout for cache efficiency
  • Core pinning across P/E cores for predictable latency
  • Modular I/O · tiling · IPC · kernel separation
C++IPCSHM
View source

MiniGit

C++ · Systems

A Git-like version control system in modern C++ with init, commit, branch, and merge.

  • Command / Factory / Strategy / Singleton patterns
  • Hash-based content snapshots + custom serialization
  • Multi-threaded commit indexing with in-memory caching
C++STLConcurrency
View source
04 / stack

Toolkit

Languages

GoC++20PythonSQL

Low-Latency Systems

epollLock-free SPSCNUMA pinningMulticast UDPSHM / IPCCache-line alignmentBranchless hot paths

Trading Infrastructure

gRPCNATS JetStreamPrometheusCOLO streamingState machinesRMS / pre-trade risk

Markets & Domain

Equity derivativesCommodity derivativesCurrency & ratesOrder lifecycleMTM accounting

Cloud & Data

AWS (S3, multi-AZ)Apache AirflowMongoDBDocker

AI / Agents

CrewAIRAGPydanticGraph databases
05 / credentials

Certifications & recognition

SEBI NISM derivatives certifications grounding the domain modeling, plus cloud, research, and awards.

SEBI

NISM Series-VIII

Equity Derivatives Certification

SEBI

NISM Series-XIII

Common Derivatives — Equity, Currency & Interest-Rate

View credential
SEBI

NISM Series-XVI

Commodity Derivatives Certification

View credential
AWS

AWS Solutions Architect

Multi-AZ, disaster-resilient cloud architectures

View credential
Research

IEEE Xplore Publication

RAG systems using AI models & graph databases

View credential
Award

Codeathon — 1st Prize

₹1L, awarded by the CEO, company-wide

View credential
06 / contact

Let's build something
that has to be fast.

Open to conversations about trading infrastructure, low-latency systems, and quant engineering. The fastest way to reach me is below.

souvikmukherjee150@gmail.com+91 7330787625
GitHub LinkedIn

Delivered securely · no tracking