Abhishek Prakash

Building AI chips at AWS

Seattle, WA --:--

Work

Abhishek Prakash

Senior Software Engineer, Annapurna Labs, AWS

Senior Software Engineer and technical leader with 7+ years at Amazon and AWS, with a track record of building new teams, platforms and production systems from the ground up.

Connect on LinkedIn
  • Next-gen AWS AI chips

    Developing next-generation AWS AI chips and optimizing AI inference.

  • Intern to Senior in 4.5 years

    Intern to Senior SDE in 4.5 years at Amazon.

  • Founding engineer and tech lead

    Founding engineer and tech lead, helped build multiple teams from the ground up at Amazon and AWS, including AWS Security Incident Response.

  • ICPC Asia Regionals

    Competitive programmer, ACM ICPC Asia Regionals.

My work has evolved from large-scale distributed systems and ML platforms into increasingly deeper layers of the AI stack. At Amazon, I re-architected static, rule-based pricing into an ML-driven platform with A/B experimentation and built the Digital Pricing Knowledge Graph. I went on to lead the launch of AWS Security Incident Response and its agentic AI-powered investigation capabilities, then architect Amazon’s next-generation GenAI-driven customer understanding and recommendation platform. Most recently, I joined Annapurna Labs to help design next-generation AWS AI chips and optimize AI inference, using assembly-level performance analysis to guide chip design iterations.

Across these initiatives, I have defined technical strategy, led engineers across organizational boundaries, and built at the intersection of distributed systems, security, ML and GenAI, and high-performance AI infrastructure.

Toolkit

Work highlights (5)

Five systems, newest first.

AI chip design and inferenceA conceptual processor package and signal board, representing AI chip design and inference kernels. These shapes are illustrative and do not depict a proprietary hardware design or production topology.AI chip designInference kernels

  1. Annapurna Labs, AWSSenior Software Engineer

    Next-generation AWS AI chip design and inference

    At Annapurna Labs, I contribute to the design of next-generation AWS AI chips at the pre-silicon stage. I also lead the architecture and optimization of AI inference kernels, inventing techniques to maximize hardware performance.

    • Pre-silicon chip design
    • AI inference kernels
    • Inference kernel optimization
    • Hardware-software co-design
    • Technical leadership
    Read the case study: Next-generation AWS AI chip design and inference

  2. AmazonLead Engineer, New Initiative

    Built from the ground up

    GenAI customer understanding and recommendation platform

    I was the lead engineer for a new Amazon initiative, building its next-generation customer understanding and recommendation platform from the ground up. I led the architecture and technical direction and helped grow the team, using GenAI and machine learning to understand customer behavior and improve personalization for 100M+ customers.

    Recommendation system and business impact: Led the architecture and development of a multi-stage recommendation system combining traditional ML models with GenAI to improve personalization. Initial experiments showed estimated annualized gains of $50M+ in incremental sales and 10M+ units.

    • vLLM
    • LLM fine-tuning (QLoRA)
    • Distributed training
    • Recommendation systems
    • Agentic AI (AWS Strands)
    • Agentic recommendation validation
    • Causal measurement
    • Model quality and benchmarking
    Read the case study: GenAI customer understanding and recommendation platform

  3. AWSFounding Engineer and Technical Lead

    Built from the ground up

    AWS Security Incident Response

    I helped architect and build AWS Security Incident Response from the ground up, leading a team of 12+ people through its public launch at AWS re:Invent 2024. The service subsequently expanded to 15+ AWS Regions.

    Agentic investigation: Led the architecture and development of AI-powered investigation workflows that gathered and correlated evidence across AWS sources. Investigation first drafts took approximately 15 minutes, compared with 10–12 hours of manual work, with analysts reviewing and approving the findings.

    • Threat modeling
    • Application security
    • Penetration testing
    • Distributed systems
    • Identity and access governance
    • REST API design
    • Public API documentation
    • ML-based security finding triage
    Read the case study: AWS Security Incident Response

  4. Amazon Digital Pricing ResearchSoftware Development Engineer II

    Built from the ground up

    Digital Knowledge Graph

    I conceived Amazon’s Digital Pricing Knowledge Graph, wrote its PR/FAQ, pitched the idea to leadership and secured funding for its development.

    Graph architecture: Designed the graph model and built the platform using Java, Gremlin, a domain-specific language and Amazon Neptune, supporting 50M+ vertices and adoption by 5+ global business units.

    • Amazon Neptune
    • Gremlin
    • Knowledge graphs
    • Graph data modeling
    • Custom DSL
    • Boolean rule engine
    • Federated data architecture
    • Streaming data ingestion
    • Distributed systems
    Read the case study: Digital Knowledge Graph

  5. AmazonSoftware Development Engineer II

    Digital Pricing Research Platform

    I led the re-architecture of a static, rule-based pricing engine into an ML-based platform with A/B experimentation. I partnered with scientists on platform design and owned the experimentation architecture, while also building customer-facing pricing services and improving the underlying AWS infrastructure.

    Pricing-service performance and cost: Implemented architectural and AWS infrastructure improvements that reduced pricing-service latency and delivered $1.2M in annual operating cost savings.

    • ML platforms
    • A/B experimentation
    • Distributed systems
    • Low-latency services
    • High-throughput services
    • Performance optimization
    • AWS infrastructure
    Read the case study: Digital Pricing Research Platform

Experience

PromotionCurrent roleSelect a step to open its role below.

All roles

  1. 2026 – PresentAnnapurna Labs, AWSSenior Software Engineer, Inference

    Seattle, WA

    Inference performance on the next generation of AWS AI chips, joined at the pre-silicon stage.

  2. [2025 – 2026]AmazonSenior SDE, lead engineer for a new GenAI initiative

    [Seattle, WA]

    Set the platform’s multi-year roadmap on foundation models and fine-tuned LLMs, and contributed across its model-quality program, from benchmarks and training-data quality to state-aware ranking, where label contamination rather than model capacity proved the main limitation.

  3. Dec 2023 – [2025]AWS Security Incident ResponseSDE II, promoted to Senior SDE in Q1 2025

    Arlington, VA, then Seattle, WA

    AI & API Bar Raiser for the org: led the technical review of every public-facing API launch and drove the agentic AI adoption that lifted engineering PR throughput 40%.

  4. Jul 2021 – Dec 2023Amazon Digital Pricing ResearchSDE II

    Arlington, VA

    Pricing systems for Amazon’s digital businesses: first Reference Pricing, then the Digital Pricing Knowledge Graph, both detailed in Work highlights above.

  5. Jul 2020 – Jul 2021Amazon Digital Pricing ResearchSDE I

    Chennai, India

    Redesigned the eBook bundle pricing algorithm with sliding-window schedule compression, cutting fleet CPU by 40%; built the Price Correction tool that Prime Video, eBooks, Appstore and Alexa teams use to fix pricing errors.

  6. 2016 – 2020EarlierSDE intern, Amazon (2019 – 2020); B.Tech Computer Science, SASTRA University

    India

    Shipped a promotions system for Prime Video as a returning intern in Chennai; ranked 66th at the ACM ICPC Asia Regionals, Kanpur 2018; problem setter for GUVI CodeKata.