Pranay Singh Parihar

Pranay Singh Parihar

+•• ••••• •••••••••••@•••••.com

Founder · Inferno · 5.9 yrs

As a founder, I am driven by a strong sense of ambition and accountability, always striving for mastery in my endeavors. My approach is rooted in logic and deep concentration, allowing me to build with a clear, idealistic vision and a disciplined focus on achieving impactful results.

  • Open to Hyderabad · Bangalore
  • Remote, Hybrid, On-site

Ask about Pranay

Skills

Claimed

Developer Feedback AnalysisExternal System State VerificationOperation Completion VerificationPost-Mortem AnalysisIterative Problem SolvingStrategic PlanningDivide and Conquer StrategyEnd-to-End Solution OwnershipAI Platform DevelopmentDeployment Verification AutomationApplication MigrationResource Utilization

Proof of Experience

Inferno

Founder

Inferno · Jan 2026 – Present

  • • Built InfernoRT, an AI inference runtime that outperforms llama.cpp across every benchmark configuration where both runtimes successfully complete execution.
  • • Enabled workloads that fail on other runtimes due to memory constraints, including 8B-class models within a 4 GB RAM budget and 70B-class models within 16 GB RAM.
  • • Delivered higher decode throughput and lower time-to-first-token than other runtimes while preserving output correctness under our validated benchmark suite.
  • • Achieved almost 2x the imporvements in TTFT than llama.cpp
  • • Built a production benchmarking framework spanning 480 workloads to validate latency, throughput, correctness, and memory behavior against the industry baseline.
  • • Filed core IP for deterministic memory management enabling efficient local AI inference.
Insurjo

Insurjo'22 Cohort Member

Insurjo · Sep 2022 – Present

NCR Atleos

Project Lead

NCR Atleos · Mar 2023 – Dec 2025

  • Directed a 12-engineer infrastructure engineering team to architect and scale NCR's core foundation-model infrastructure platform, supporting 120+ active ML engineers and provisioning low-latency compute capabilities for over 9,400+ enterprise accounts.
  • Led a 12-engineer infrastructure team delivering a managed LoRA and DPO fine-tuning pipeline adopted by 1,200+ enterprise customers in its first two quarters, driving $34M in incremental ARR.
  • Designed and deployed a 2,000 GPU large-scale training cluster utilizing dedicated InfiniBand interconnect networks, compressing the distributed training timeline of a 70B parameter model from 3+ weeks down to 4 days.
  • Architected multi-tenant vLLM inference serving engines running speculative decoding and FP8 KV-cache quantization, lifting overall framework throughput 3.1x versus the legacy Triton stack.
  • Slashed per-token inference serving costs by 62% through aggressive KV-cache optimization, smarter batching windows, and model weight quantization profiles.
  • Built an autoscaling LLM inference serving platform handling 10M+ daily production requests with 99.95% high availability using Kubernetes orchestration and Triton Inference Server.
  • Optimized production serving paths to deliver an ultra-low p95 latency benchmark of 340 ms while under a sustained load of 900 QPS distributed across 3 core operating regions.
  • Engineered a high-throughput Python and FastAPI serving layer fronted by Azure Kubernetes Service (AKS), maintaining a strict p95 latency of 410 ms under peak traffic conditions.
  • Recaptured $3.2M in annual cloud infrastructure spend by designing automated spot instance preemption handlers and stateful checkpoint optimizations across deep learning jobs.
  • Developed custom cluster job scheduling algorithms that maximized hardware efficiency, successfully lifting baseline aggregate cluster utilization from 58% to 87%.
  • Orchestrated multi-tenant Kubernetes infrastructure on AWS EKS to securely isolate resource limits, namespaces, and compute boundaries for 120+ active ML engineers.
  • Implemented deep-stack NVIDIA GPU monitoring and quota management with Prometheus, Grafana, and Karpenter, expanding average GPU utilization across the org by 31%.
  • Eliminated severe training data-starvation bottlenecks by building a zero-copy data pipeline utilizing NVIDIA DALI and targeted NFS v4 transport optimizations.
  • Remediated data ingestion limits that were previously capping vision model GPU training utilization at a stalled 54%, unlocking peak hardware processing capacity.
  • Standardized declarative infrastructure patterns via Terraform and Helm, cutting GPU environment provisioning times from 2 days to under 30 minutes with a 99.2% successful release rate across Python and Go automation workflows.
NCR Corporation

DevOps Engineer

NCR Corporation · Dec 2021 – Mar 2023

  • Built GitHub Actions and Jenkins pipelines for Terraform, Ansible and serverless changes, establishing reusable delivery patterns transferable across cloud environments.
  • Developed automated monitoring, alerting and remediation workflows with Terraform, Ansible, event-driven functions and Slack/email notification integrations.
  • Automated dynamic resource configuration and application deployment through Ansible inventories, reducing reliance on manual environment changes.
  • Supported production cloud-security incidents with policy-as-code, audit logging and automated response practices.
  • Created reusable Infrastructure-as-Code modules and playbooks that reduced configuration drift by 35% and manual errors by 40% across environments.
IbexLabs

DevOps Engineer

IbexLabs · Nov 2021 – Dec 2021

  • Automated host monitoring and environment-specific container delivery with Ansible and parameterized GitHub workflows.
twimbit

DevOps Engineer

twimbit · Sep 2021 – Oct 2021

  • 1.
  • Setup CloudFront along with Application Load Balancers for ECS
  • 2.
  • Setup Task definitions for a Caddy server and static website
  • 3.
  • Write Nginx configuration to reverse proxy traffic within the cluster and to serve SSL certificate on the go with the help of OpenResty Framework
  • 4.
  • Managing the entire infrastructure using terraform.
  • Setup Lambda function for database failover and automate the process using step functions.
  • 5.
  • Setup A records and CNAME in Route 53.
Neudesic

DevOps Engineer

Neudesic · Mar 2021 – Aug 2021

  • Standardized Windows Server provisioning with reusable Packer images and Terraform/Bicep-managed Azure VM Scale Sets, reducing configuration drift across environments.
  • Resolved Azure provisioning and networking issues by codifying repeatable Infrastructure-as-Code fixes instead of manual environment repair.
  • Recognized internally as a Terraform subject-matter expert for Microsoft collaboration based on reusable cloud-automation practices.
Tivona Global

DevOps Intern

Tivona Global · Oct 2020 – Feb 2021

  • Automated continuous deployment with Bash and provisioned AWS infrastructure from scratch using Terraform, establishing a repeatable cloud-delivery foundation.
  • Implemented Terraform Cloud Sentinel policies and dynamic Terraform blocks to enforce policy-as-code and create multiple resources from reusable complex values.
  • Created IAM policies and group-based access controls; developed Boto3 Lambda automation to attach IAM policies and record CloudTrail activity in DynamoDB.

SWE Virtual Intern

JPMorgan Chase & Co. · May 2020 – May 2020

  • 1.
  • I was responsible for optimizing and troubleshooting a website.
  • 2.
  • I had to get familiar with JPMorgan Chase frameworks and tools.
  • 3.
  • Understanding the concepts of git was also part of the job.
  • 4.
  • Display data visually for traders (Trader's Dashboard)
  • 5.
  • Use JP Morgan's perspective framework.
Accenture

Australia Discovery Program

Accenture · May 2020 – May 2020

  • 1.Set Project Priorities
  • 2.Assemble a plan
  • 3.User Journey Redesign
  • 4.Outcomes Analysis
  • 5.Fix the errors
  • 6.Prioritisation & Impact Assessment
Deloitte

Technical Consultant [Virtual Internship]

Deloitte · May 2020 – May 2020

  • 1.Understanding Cloud Computing
  • 2.Cloud Feasibility Assessment
  • 3.Cloud Readiness Assessment
  • 4.Client Discovery
  • 5.Design a Business Case
  • 6.Considerations For Mobilisation
  • 7.Define the project approach
  • 8.Conduct a market scan
  • 9.Further analysis & solution presentation

Proof-of-Work

No completed sessions yet. Anything Pranay finishes shows up here with the evidence behind it.

Proof of Education

Bachelor of Technology, Electrical and Electronics Engineering

VIT University, Chennai · 2017 — 2021

A Proof CV — one profile, kept current, with the proof attached. Back to the portal