Backend, Distributed Systems & AI Platforms | Ex - BookMyShow, Contentstack

Bengaluru

Mumbai, Maharashtra, India

India

Mumbai, Maharashtra, India · Hybrid

Mumbai, Maharashtra, India · Hybrid








Bachelor of Engineering - BE, Information Technology
MCT's Rajiv Gandhi Institute Of Technology
Join Soumik and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal BossBackend, Distributed Systems & AI Platforms | Ex - BookMyShow, Contentstack
Backend SDE-II at BookMyShow, building AI platforms on top of distributed systems. Most of my recent work sits at the intersection of LLMs and backend infra: a LangGraph + Chainlit multi-agent platform that wires 4 MCPs into one UI (25+ agents across a 200-engineer org, 50% faster API and data discovery), schema-aware LLM-to-Elasticsearch query generation that cut query errors 70%, and end-to-end LLM observability via OpenTelemetry to Coralogix. Before AI infra, I owned core backend work at BMS. Led the Node.js to Go migration of our transactional comms stack (SQS + RabbitMQ, 2K msgs/min), owned the Fault Tolerance service on Cloudflare Workers that kept pages alive through Coldplay, ICC World Cup, and Travis Scott, and architected an S3 Glacier invoice pipeline projecting $146K in storage savings. Started at Contentstack on high-throughput logging at 10M+ logs/day, where I fixed a Kafka consumer GC pressure bug, built a 100% delivery-guarantee layer, and rolled out tiered Hot-Warm-Cold Elasticsearch storage. ACM-ICPC Amritapuri Finalist (2021, 2022), Codeforces Specialist, CodeChef 4-Star. Still useful when I'm reasoning about systems and performance. What I'm curious about next: backend system design, the kind of complex trade-offs that take days to reason through, and the problem-solving that scale forces. AI infra sits on top of that for me, where I'm interested in agentic systems that actually work in production and LLM cost and latency engineering. Open to backend, platform, and AI-infra roles. If you're building something interesting, drop me a line. Tech: Go · Python · TypeScript · LangGraph · LangChain · MCP · FastAPI · Cloudflare Workers · AWS · Kafka · MongoDB · Elasticsearch · OpenTelemetry
Mumbai Metropolitan Region
claude code
claude
codex
cursor
opencode
Backend Engineer at Xflow, Jul 2026 - Present · 3 mo, Bengaluru
Software Development Engineer - II at BookMyShow, Apr 2026 - Jul 2026 · 4 mos, Mumbai, Maharashtra, India, • Architected a 4-MCP multi-agent platform (Python, LangGraph, Chainlit) that serves 25+ agents to a 200-engineer org, cutting API and data discovery time 50% (15 eng-hrs/week saved). • Built the OpenAPI discovery MCP (FastAPI) with hybrid retrieval combining TF-IDF, cosine similarity, and rapidfuzz tie-breaking; traced end-to-end via OpenTelemetry into Coralogix. • Designed the EventStore MCP, a schema-aware LLM-to-Elasticsearch DSL with validation that cut query errors 70%; 1 NL tool + 13 structured analytics tools, with a static-query fallback for cost control. • Shipped the Database Intelligence MCP: 10-tool server + offline schema exporter across MongoDB, Elasticsearch, PostgreSQL, MariaDB, and MSSQL, with cross-DB relationship inference via field-name and type heuristics.
Software Development Engineer - I at BookMyShow, Dec 2024 - Apr 2026 · 1 yr 5 mos, India, • Led architecture for the Node.js to Go migration of BMS’ transactional comms (email/SMS/OTP/WhatsApp): pluggable SQS + RabbitMQ queue layer, adaptive polling, 3-tier retry/DLQ topology at 2K msgs/min. • Owned the Fault Tolerance service on Cloudflare Workers KV, with region-batched URL refresh through a Go platform API (MongoDB audit, distributed locking); kept pages alive through Coldplay, ICC World Cup, and Travis Scott. • Architected an invoice retention pipeline on S3 Glacier Instant Retrieval (chosen over Deep Archive for sub-ms audit retrieval); meets 8-year regulatory compliance with $146K projected in storage savings. • Built automated OpenAPI spec generation from live traffic: PII-masked TCP dumps flow through an n8n workflow that correlates K8s endpoints and uses Claude to produce specs, auto-committed as PRs to Backstage. • Automated the company-wide appraisal cycle for 50 managers in Google Apps Script, cutting coordination overhead 80% and collapsing a multi-day process into hours.
Associate Software Engineer at Contentstack, Aug 2023 - Nov 2024 · 1 yr 4 mos, Mumbai, Maharashtra, India · Hybrid, • Managed Applogs-as-a-Service at 10M+ logs/day; tracked down and fixed a GC pressure issue from unbounded byte slice retention in long-lived Kafka consumer goroutines, eliminating weekly OOM pod restarts. • Built an end-to-end delivery guarantee layer for the logging pipeline, reaching 100% delivery rate with eventual consistency across the distributed architecture. • Rolled out tiered Hot-Warm-Cold storage in Elasticsearch with lifecycle policies, template mappings, and alias management; improved SLO compliance by 1.6% and cut operational costs by 9.2%. • Resolved a critical Kafka partitioner bottleneck causing single-partition hotspotting, bringing message processing latency down 13.6%.
Associate Software Engineer Intern at Contentstack, Jan 2023 - Jul 2023 · 7 mos, Mumbai, Maharashtra, India · Hybrid, Handling and scaling backend microservices being part of the Platform team.
Software Development Engineer Intern at Ombré, Jun 2021 - Dec 2021 · 7 mos, - Developed an Image Compression Cloud function in NodeJS which helped in reducing the image sizes for our mobile application by 90% hence reducing loading time by 57% for our Dashboard (Homepage) images. - Migration of codebase from Flutter V1 to V2 hence improving the stability of mobile applications by introducing null-safety across platforms and devices. - Migration of User-Media collection (Audios, Videos, User Highlights, User Links) hence minimizing the cost for user queries by 4 times from the database. - Remodeled the Dashboard UI, and Revamped the User-Profile page UI, which decreased the page navigation by removing a page, hence improving the UX performance.
Flutter developer at Bottom Street, Mar 2021 - May 2021 · 3 mos, - Integrated REST API, Firebase for authentication, storage, testing purposes, applied state-management (Riverpod) using Flutter. - Designed and implemented a feature for showing trending search and recent search which helped in the retention of 23% of the users. - Implemented Deep Linking into the mobile application. - Worked on Payment portal using Razorpay (Flutter).
Bachelor of Engineering - BE, Information Technology, MCT's Rajiv Gandhi Institute Of Technology
github: https://github.com/Soumik43
leetcode: https://leetcode.com/Soumik43
codeforces: https://codeforces.com/profile/soumo943