Product - Tiyaro AI APIs: Vast types of ML models - LLMs, Embeddings, Image Generation, Time-Series, etc., readily available for production use with just a few clicks.• System, Backend + Infra services to enable scalable, reliable, performant Inference APIs with average GPU memory util of 75%.• Onboarding & release new OSS models from HuggingFace to Prod.• Founding Engineer. First hire in Berlin.Product - Tiyaro DeepQuery• Infra services to deliver business value using LLMs & Generative AI via Retrieval Augmented Generation (RAG) & AI Agents.• Datapipeline & DAG workflows infra for building feature artifacts to be used by RAG & AI Agents.Tech Stack--------------• Python, Go• ML Model Frameworks - Transformers, PyTorch, TensorFlow• OpenSearch, Qdrant, Postgres, MLFlow, HuggingFace, NVIDIA Triton, DVC, S3• AWS, GCP, K8s, Docker• DevOps - Infrastructure as Code - CloudFormation Templates & Helm Charts, CICD - AWS CodeBuild, Argo CD & Workflows, Makefile

As part of Platform Engineering, built & maintained systems to meet growth plans for Careem handling 20M req/day for 50M users.• End-End ownership of Pairwise Pseudonymous Identifiers (PPID) based AccessTokens feature for 3rd party apps to enable Careem “SuperApp”.• Led efforts to improve scalability by 10X (50k RPM) & decreased DB load using several performance optimization techniques like caching, decoupling joins, optimized query execution times & sharding. • Led efforts to migrate and shutdown legacy less-secure stateful access token systems - savings of >300k per year.Tech Stack--------------• Java 8, SpringBoot, JUnit, Mockito, Maven; Python, Kotlin• Amazon Web Services - S3, SQS, DynamoDB, Lambda, Kinesis Stream, IAM, etc.,• DevOps - Infrastructure as Code using Terraform, CICD - Github Actions & Jenkins, Makefile, Earthly• Alerts, Monitoring, Performance Debugging

• OAuth2–OIDC based Login, 2FA, SingleDeviceSession, custom-grants, etc., features to enable “SignIn with Careem”.• Wrote design documents (RFCs) for several features. Mentoring & leading 3 engineers in developing features. Collaborated across teams for integration, onboarding, on-call rotations & as well as incident handling.• Designed & implemented several fraud prevention strategies based on proactive threat models for new features and, reactively for newer sophisticated attacks while balancing costs and UX for legitimate users.

• Communication Services - Template Editor, Config Management, Email, SMS, Push Notifications• Worked on the design and implementation of Communication microservices, that is responsible for all communications sent through different channels to Drivers and Customers, serving on average ~12M req/day.• Designed and developed a scalable event driven system capable of partial updates using Kinesis, DAX & DynamoDB.

Worked closely with the Office of the CTO – Internet-of-Things in developing VMware's open-source IoT Agent and SDK - LIOTA. Implemented the following:• Developed support for MQTT as transport layer for Device <--> Edge <--> Cloud message passing. • AMQP, Bluetooth Low Energy (BLE) protocols as transports, simple Edge Filtering capabilities.• Involved in designing and implementation of low latency, scalable system with message-broker architecture.• Designed and developed POCs for edge intelligence feature. Movement of ML models from Cloud <---> Edge aligning with VMware’s IoT strategy.Tech Stack--------------• Java, Python

Did my GSoC project with WSO2 (www.wso2.org). Developed a high performance, lightweight, low latency HTTP Load Balancer built on top of Netty based non-blocking IO and Disruptor (ring-buffer) architecture.• Support for different Load Balancing Algorithms (RR, Weighted RR, Random)• Session Persistence through methods like Application or LB Cookies & Consistent Hashing on Client IPs• Health Checking and Redirection - Active & Passive health checking• Performance benchmark using Apache Bench, results greater than Nginx (Open source).• Performance issues debugging using JFR – Java Flight Recorder.• https://summerofcode.withgoogle.com/archive/2016/projects/5891422477615104/ Tech Stack---------------• Java, OSGi bundles• ANTLR 4 for DSL

Interned here after my 4th semester as an Android Application Developer for a Wireless Health Care Solution (Wearable wireless sensor based system for emerging market opportunities).Tech Stack--------------• Android (Native), Java, XML