CV

Contact Information

Name Akash Srivastava
Professional Title Senior Engineer / Tech Lead · Distributed Systems · Cloud Infrastructure
Email akash.s1684@gmail.com

Professional Summary

Senior Engineer / Tech Lead with 10 years of experience specializing in AI infrastructure, distributed systems, and real-time platform engineering. Sets technical direction and drives architecture and delivery of AWS real-time voice and ML inference platforms, leading large cross-functional initiatives across team boundaries. Extensive experience building low-latency, highly available services at global scale, including multi-region systems supporting 100k+ concurrent users and demanding production workloads.

Experience

  • 2021 - present

    Seattle, WA

    Senior Software Development Engineer / Technical Lead
    Amazon Web Services (AWS)
    • Technical lead in the real-time Voice and ML inference domain; set technical direction, led architecture reviews, defined long-term platform vision, and balanced iterative delivery against immediate business results.
    • Architected and operated a globally distributed real-time inference platform supporting 100k+ concurrent users, owning service architecture, scaling strategy, fault tolerance, deployment automation, and latency optimization across multiple AWS services.
    • Re-architected legacy real-time streaming services, establishing the core streaming infrastructure underpinning a large-scale omni-channel contact center platform with 99.9%+ availability.
    • Designed and implemented a multi-region inference platform integrating external TTS providers with automated failover, traffic routing, observability, and latency-aware request handling, achieving 99.99% availability.
    • Designed a GPU benchmarking framework across multiple instance types to evaluate performance/cost tradeoffs for TTS inference workloads; analysis directly informed infrastructure selection and reduced serving costs for the production stack.
    • Architected a distributed video transcoding pipeline — segment-based ingestion, per-contact demuxing, audio/video muxing and re-encoding into compressed MPEG-4, with KMS-encrypted delivery to customer S3 at scale.
    • Built a reproducible SageMaker-based ML training pipeline, replacing manual workflows and reducing model iteration time by 85% while improving experiment tracking and reproducibility.
    • Drove noise suppression and speech quality improvements across telephony and WebRTC deployments, improving average DNS-MOS score by 20%.
    • Hiring Bar Raiser passionate about mentoring and hiring; conducted 300+ interviews and influenced hiring decisions across multiple teams and roles.
  • 2017 - 2021

    Hyderabad, India

    Software Development Engineer I → II → III
    Amazon — Marketplace & Seller Platform
    • Progressed from SDE-I to SDE-III building large-scale seller productivity and growth systems on Amazon’s service-oriented architecture, including the Seller VAT Calculation Services Engine processing 250M+ monthly events across 300k+ sellers globally.
    • Owned end-to-end design, scaling, and delivery of high-throughput seller workflows; drove cross-team technical leadership and mentored engineers through significant org growth.
    • Led architecture reviews and long-term re-architecture initiatives to evolve extendable, globally scalable services, reducing operational toil and improving system resilience across Amazon Marketplace.

Skills

Systems: Distributed systems, Kubernetes/EKS, real-time streaming, event-driven architecture, domain-driven design, low-latency systems, WebSocket, HTTP/2 streaming, gRPC, REST, WebRTC
AI / ML: LLM inference, ASR, Text-to-Speech (TTS), Speech-to-Speech (S2S), ML model serving, inference optimization, GPU benchmarking, FFmpeg, video/audio transcoding, SageMaker pipelines, noise suppression
Cloud / Infra: AWS (Kinesis, ECS, Redis, SageMaker, Elasticsearch, Docker, DynamoDB), multi-region fault-tolerant deployments
Languages / Data: Java, C++, Python, relational databases, NoSQL, real-time analytics pipelines

Publications

  • 2018
    A Deadlock-Free Lock-Based Synchronization for GPUs
    Concurrency and Computation: Practice and Experience, Wiley

    A lock-based synchronization primitive for GPU architectures that is provably free of deadlock.

  • 2015
    Performance Comparison of Various Particle Swarm Optimizers in DWT-SVD Watermarking for RGB Images
    ACM ICCCT

    A comparative study of PSO variants applied to optimize DWT-SVD image watermarking, improving watermark fitness by 22% over the conventional scheme.

Education

  • -

    Bachelor of Technology
    National Institute of Technology (NIT), Patna
    Computer Science