Sriram Kumar
Engineering Leader & Platform ArchitectI am an engineering leader with 14 years of experience building multi-tenant SaaS and AI-native platforms. My core focus is solving complex scaling bottlenecks and transitioning legacy monolithic systems into resilient, event-driven architectures. As a hands-on 'Builder-Leader', I've scaled engineering organizations to 40+ people and delivered platforms that process massive transaction volumes for highly regulated BFSI and B2B sectors. I believe in pragmatic technical choices, cost-disciplined infrastructure, and building high-trust teams that deliver predictably.
Platforms I have built and scaled have been deployed across these organizations:
My Operating Playbook
1. Resilient & Multi-Tenant Architecture
I design for 10x burst traffic and strict tenant isolation. By transitioning legacy monoliths to event-driven microservices with backpressure controls, I ensure enterprise systems degrade gracefully rather than failing under load.
2. Pragmatic, AI-Native Systems
I move beyond wrappers to embed AI into core B2B workflows. I focus on architecting deterministic guardrails around non-deterministic LLMs, optimizing inference latency, and running real-time interaction pipelines securely.
3. The Builder-Leader Culture
I build high-trust, high-ownership engineering organizations. I drive delivery predictability by enforcing automated CI/CD hygiene, progressive delivery gates, and blameless post-mortem rituals that raise the technical bar.
Execution & Architecture
A deeper look into hands-on system design, capacity planning, codebase modernization, and architectural trade-offs at scale.
Event-Driven Modernization for Burst Scale
The Challenge
Legacy monolithic processing failing under unpredictable enterprise burst traffic.
Architecture & Execution
Spearheaded the migration to an asynchronous, Kafka-backed event-driven topology with strict backpressure controls. Stayed deeply hands-on in topology decisions and production load-testing.
Business ROI
Achieved zero downtime during 10x traffic spikes, successfully absorbing burst events while delivering automation that reduced end-client agent workload by 40%.
Enterprise AI Inference at Scale
The Challenge
Multi-turn latency and high compute costs making real-time generative AI unviable for enterprise adoption.
Architecture & Execution
Designed an asynchronous, event-driven interaction pipeline for STT, LLM inference, and TTS. Implemented strict evaluation guardrails to manage non-determinism.
Business ROI
Resolved critical production bottlenecks for real-time avatars while driving down compute inference costs by ~40% through intelligent workload isolation and autoscaling.
Delivery Predictability in Regulated Environments
The Challenge
Multi-week deployment cycles bottlenecking feature delivery and customer onboarding for Tier-1 banking clients.
Architecture & Execution
Engineered robust CI/CD pipelines, automated testing protocols, and progressive rollout gates (canary deployments) aligned with strict SOC-2 constraints.
Business ROI
Reduced code-to-production time from 2+ weeks to under 2 days, increasing deployment frequency by 3x and accelerating enterprise onboarding timelines by 30%.
Career Timeline
A history of execution, leading teams, and scaling systems.
Director of Engineering
Problem Space: Building the foundational engineering organization and platform architecture from the ground up to support AI-native real-time interactions.
Key Decisions: Designed an event-driven real-time interaction pipeline for STT, LLM inference, and TTS (rejected synchronous request-response patterns after proving latency issues).
Challenges: Managing non-determinism in LLM outputs, optimizing inference latency, and creating evaluation guardrails for production-ready AI pipelines.
Outcomes: Reduced code-to-production from 2+ weeks to under 2 days. Maintained 99.9% uptime while lowering compute costs by ~40%.
Associate Director of Engineering
Problem Space: Modernizing legacy monolithic enterprise conversational products (WhatsApp, Web, RCS) for Retail and BFSI customers running at high scale.
Key Decisions: Pioneered migration from monolithic message processing to asynchronous event-driven architecture using Kafka. Led hybrid build-vs-buy strategy for NLP and messaging components.
Challenges: Migrating high-throughput monolithic services without downtime, load testing burst traffic events, and creating backpressure controls.
Outcomes: Reduced agent workload by 40% and improved end-user session retention by 65%. Delivery automation directly influenced enterprise upsells.
Director of Engineering
Problem Space: Scaling delivery and architecture for digital banking assistant platforms used by Tier-1 banks across India, the US, and the Middle East.
Key Decisions: Introduced a pod-based topology aligned to banking product streams. Made complex build-vs-buy choices across NLU and dialogue orchestration based on enterprise readiness requirements.
Challenges: Engineering highly reliable systems that meet strict banking compliance constraints, and managing critical production escalations for Tier-1 deployments.
Outcomes: Reduced customer onboarding timelines by 30% through platform and process improvements. Hired 8 engineers and developed 3 into first-time managers.
Engineering Manager
Problem Space: Improving deployment velocity and platform robustness for conversational AI systems.
Key Decisions: Built delivery risk and compliance practices that evolved into org-wide engineering standards.
Challenges: Automating testing protocols and optimizing CI/CD pipelines to ensure predictable release cycles.
Outcomes: Improved roadmap adherence to 98% and increased deployment frequency by 3x. Supported 30% YoY platform usage growth.
Lead Engineer
Problem Space: Designing core conversational AI foundations for enterprise banking.
Key Decisions: Architected early services allowing safe multi-tenant deployments.
Challenges: Establishing a flexible API gateway to interact seamlessly with varying legacy core banking systems.
Outcomes: Enabled successful early enterprise go-lives for major financial institutions.
Lead Engineer
Problem Space: Building robust global educational technology platforms.
Key Decisions: Technical stack stabilization and cloud deployment approaches.
Challenges: Scaling an ed-tech platform across 25+ countries and supporting 20K+ educators smoothly.
Outcomes: Successfully managed scale resulting in high user engagement globally.
Software Engineer
Problem Space: Developing digital banking applications.
Key Decisions: Operated as a core individual contributor, focusing on the technical execution and delivery of secure backend modules for mobile banking platforms.
Challenges: Ensuring secure transactional capabilities for APAC and Middle East banking clients.
Outcomes: Delivered enterprise digital banking modules.