Karthik Gooli — Senior Staff Engineer

Professional Experience

Fifteen years across every stage of company — large, early, growth, pre-IPO. Kernel and file systems at NetApp, SOCs at NVIDIA, distributed systems and data at Uber, finance at scale at Fisdom, data at scale at Peritus, security at scale at Foxpass. A generalist by nature, with a strong bias towards action. The domain and context keep changing. The loop never does: identify bottlenecks/challenges, respond to company and customer needs, prototype it, build it right, scale, and keep it less ops heavy.

foxpass.com Current

March 2021 — till now

Sr Staff Engineer · Headquartered in San Francisco

YC-backed startup delivering server and WiFi authentication to SMBs — acquired by Splashtop (unicorn) in 2023. Key contributor to driving Foxpass to multiple millions in ARR.
Architect of three core product offerings — RADIUS, LDAP, and PKI — deployed globally as horizontally scalable, multi-tenant services, leading a team of engineers across US and Asia including China and India. Track record of building products from the ground up that open new revenue streams and are consistently praised by customers for ease of deployment, use, and reliability.
Own the end-to-end vision for Foxpass's authentication ecosystem — comprising the core authentication platforms (RADIUS, LDAP, PKI) that form the backbone of client authentication, and the product and administration layer with a rich set of customer-configurable knobs and policies that dynamically govern how the authentication platform behaves for every customer, dictated by the policies they set for their resources and users.
Own the design, scaling, and efficiency of Foxpass's homegrown multi-tenant LDAP cluster — built on async Twisted, serving 20–25k QPS with 150k persistent TCP connections across hundreds of customers.
Implemented context-aware search indices that reduced LDAP search latency several-fold and cut CPU utilization by 30–40%; resolved key scalability bottlenecks that unlocked seamless growth, helped land crucial big-ticket customers, and enabled scaling in a more cost-effective way.
Architected and built a clustered LDAP system with partitioning from the ground up — evaluated bottlenecks, prototyped, and implemented the full solution including changes to open-source libraries. Reduced per-node memory by 40% and DB reads by 75%. Running strong in production serving several hundreds of customers.
GitHub Blog
Re-architected Foxpass's global RADIUS infrastructure end-to-end — reimagined the platform to operate behind a global accelerator with lightweight PoPs deployed worldwide, routing authentication requests to the nearest instance and significantly reducing latency across geographies.
Designed and built passwordless WiFi authentication for Chromebooks as a horizontally scalable, multi-tenant service — replaced Google's non-scalable, ops-heavy GCCC with a solution that spins up with a single click and zero manual intervention. Consistently praised by customers for its ease of deployment; best-in-class in the industry, onboarded several customers, opening a significant new revenue stream, loved by our customers.
Pioneered the concept of a virtual RADIUS server — designed and built from scratch in async Python, allowing each customer to operate behind their own custom port. Deployed as a service at scale leveraging the shared authentication platform, serving several hundreds of customers.
Designed and built a RadSec (RADIUS over TLS) server from the ground up as a horizontally scalable, multi-tenant service, capable of handling 10k+ concurrent connections on a single instance. Deployed at scale globally, overcoming the shortcomings while still leveraging the existing authentication platform.
Blog
Designed and implemented eduroam — successfully authenticates home students and seamlessly routes visiting students to their own institution's authentication servers, enabling universities and institutions worldwide to leverage Foxpass for secure network access.
Blog
Implemented SNI (Server Name Indication) on RadSec servers — allows customers to configure their own custom DNS and provision certificates specific to that DNS, giving each customer a dedicated and secure TLS endpoint.
Architected a new PKI product offering leveraging the existing certificate generation infrastructure — opened up a significant new revenue stream.
Product page
Designed/Implemented the backend for Global lookup feature which helps users access any of our stack (US/EU/AU) seamlessly. Helps new users to be redirected to the platform closer to them and of their choice.
See it live
Contributed to Golang/python/C open source cryptography library that's used by millions of users around the globe.
pkcs7 PyScep asn1crypto
Introduced DocumentDb for ldap logging purposes. Replaced old non performant logging backend with a complete new backend which made query results to appear 10x faster and with no visible latency from the front end.
Created a new python library called 'TTLStore' and available for general use. This is also used in our production for different use cases.
PyPI
Responsible to ensure code quality and successful monthly releases.
Tech used: Postgres, S3, Athena, ECS, Docker, Python, datadog, scalyr, rollbar, Django for our console/admin needs.

Peritus.ai

Jan'18 — Feb'2021

Founding Engineer / Tech Lead · headquartered in Palo Alto

2nd engineer in peritus.ai (Total strength of the company currently: 15); company acquired by Essenvia
Patent US10679133B1
Google Patents
Peritus.ai provides machine learning insights and automation to scale community intelligence and provide recommendations.
Designed and implemented a horizontally scalable multi-tenant continuous data ingestion platform from scratch. Grew the team to 5 engineers. This used s3/google storage, athena/big query and postgres as part of the solution. The data platform is responsible for getting data in realtime from 300+ customer community websites including stackoverflow.com, github.com, discourse.org
Created libraries to read data, create tables/views for Athena and GCP programmatically for the data sitting in s3/google storage. Automated (made it repeatable) the whole process of normalizing data from various sources and exposing it through one final view.
The data exposed through the final view served as an input for the ML/IR pipeline steps.
The whole data ingestion platform was deployed on Kubernetes.
Conceptualised, designed and implemented Rules engine. From wire framing the front end design to leading and shipping the entire project. Implemented the backend completely. This allowed authors to author rules that can be applied to terabytes of machine data.
Implemented a search solution using 100% open source software (apache license) SourceGraph which lets people search code across thousands of repositories. Productionized and integrated it with our existing search solution.
Setup a robust deployment process for processing the data through our ML pipelines so that anyone could do it.
Contributed to several components of the code base from Data platform to infrastructure to backend.
Conducted several experiments to evaluate and scale Dgraph. Dgraph was used as a knowledge graph to store machine configs to answer some important questions.
Conducted several experiments to scale solr and came up with a strategy to scale solr
Evaluated kubeflow as a platform for model training and deployment.
Introduced/wrote libraries in our code base so that mongodb could be used as a feature store to store and use the features extracted as part of the ML pipeline step for model training.
Introduced Redshift to process large volumes of data sitting in s3.
Mentored interns, entry and senior level engineers.
Java and Python are the languages used.
Kubernetes was used for the all deployment purposes
Have hands on experience on writing docker files for containerized deployment in kubernetes. Experience in docker-compose/helm charts.

Fisdom (fisdom.com)

Jan'17 — Jan'18

AVP, Engineering

Fisdom raised ~46 million USD and were acquired recently in an all cash deal for $150 million by Groww.
Built back-end for the wealth management app
Setup engineering best practices (code reviews, test setup (jenkins, jmeter), alerting, monitoring, sentry setup)
Introduced MySQL based on their data needs which decreased latency of serving the data by several folds [3s to 400ms] and rewrote components due to this change.
Wrote the whole back-end for the do it yourself feature which has seen increased engagement and investments since the launch.
Blog
Also launched top trending funds based on the demand
Introduced graphs in the mobile apps serving from the back-end using high-charts.
Automated payments of sub broker commissions end to end without any manual intervention for raising invoices
Prototyping efforts:
Used computer vision library, OpenCV, to crop images. For reducing work load on ops team
Prototyped WhatsApp based customer on-boarding using the library yowsup.
GitHub
Worked on building an AI chat bot using Google's api.ai. This is to automate the repeated customer questions.
Before narrowing down on this, prototyped several others (wit.ai, luis.ai, motion.ai)
Google app engine is used for all the infrastructural needs
Mentoring engineers to deliver high quality product Languages used: Python, Javascript

Uber, San-Francisco

Jan'15 — Feb'16

Software Engineer 2

One of the early engineers to join the effort of building the infrastructure team at Uber
2nd engineer to join the effort of building Kafka platform, helped grow team from 2 -20 in a span of 6 months
Built and scaled the logging/events infrastructure using Kafka (messaging bus) ground up. Built it across US and China and made them inter dc aware
Built 4 clusters each of 16 nodes across US and china data center, each cluster getting a traffic of 30-40k qps
Built monitoring and alerting for the infrastructure
Built logging libraries in Golang to produce messages to Kafka used by engineers across the company
Migrated the complex kafka 7 architecture to kafka 8 architecture
Wrote tools to consume messages from kafka in a restricted manner between given two timestamps, which is used companywide to consume messages from kafka
Wrote many other tools to help in operational needs of Kafka distributed systems
Got promoted within 6 months
Languages used: GoLang and Python

Uber, Bangalore

Feb'16 — Jan'17

Software Engineer 2

One of the first engineers to join the Bangalore Uber team
Helped Uber setup engineering in Bangalore. Helped grow from 0-20 engineers
Designed, implemented and successfully shipped the vehicle leasing product ground up (Platform to provide leasing opportunities to drivers)
Introduced Kafka, samza, spark in Uber Bangalore to solve the leasing problem in India
Designed and implemented a framework to scale vehicle leasing service using SWIM protocol (gossip variant) and sevnup.
GitHub
Wrote technical specs to evaluate GPS vendors.
Vehicle leasing service was built using Golang

NVIDIA, Santa Clara

May '14 — Aug'14

System Software Intern in Android Kernel Development Team

Designed and Implemented a framework in Android kernel to analyze and co-relate kernel events with architectural events. Generated visualizations using the data generated from the framework, identified hot-spots and pointed out architectural anomalies when co-related with the kernel events

Graduate Student Researcher

July '13 — Dec '14

Systems Research Group in UIUC

Built infrastructure to analyze heap growth pattern of an application. Memory footprint of an OS running on a VM is extracted using virtual machine introspection tools provided by Virtual-box. Heap dumps of the applications from the memory image is extracted using a tool called volatility. Raw heap is then reconstructed to visualize the data structures. This reconstruction helps us understand the transformations that the heap undergoes at various points of time.
GitHub Research page Poster 2013 Poster 2014

NetApp India, Bangalore

July '09 — July '13

Engineer

Development experience in kernel and distributed file system called WAFL (write anywhere file layout).
Authored 4 technical invention papers and have been awarded Patent incentive award for all of them.
Have worked on file system recovery issues and have provided patches aiding file system recovery.
Have worked extensively in SAN area. Have a good exposure in SCSI and FC protocol. Have provided root cause analysis for interoperability issues and have provided fix for the same.
Delivered tech talks, mentored interns, have been awarded for building a tool called i-sense which saved a lot of engineering resources. This tool automated the file system damage assessment and recovery time. Languages used: C

Karthik
Gooli

Professional Experience

Education

Projects executed at UIUC

KarthikGooli

Professional Experience

Education

Projects executed at UIUC

Karthik
Gooli