Headline

Software Engineer 2 @ Electronic Arts | python, pytorch, CUDA, pyspark, MCP, langchain, langgraph, agents, LLM serving, VLLM, fine tuning models, Mlops, kubernetes, keda, kserve, k8

Location

Gurugram

AI stack

claude code

hugging face

kimi

langchain

langgraph

langsmith

Work experience

Senior Technical Lead at OakNorth, Jul 2026 - Present · 2 mo, Gurugram

Software Engineer 2 | AI & MLOps at Electronic Arts (EA), Sep 2022 - Present · 4 yrs 1 mo, India, - Developing one click deployment of Vector Database: Pinecone in GenAI gateway( UI – next.js & backend- Gitlab Pipelines) running on AWS Kubernetes pods. - Implemented custom nodes in ComfyUI for stable diffusion, using CUDA programming to enhance GPU accelerated workflows and model efficiency for UI/UX teams. - Implemented RAG pipeline connecting VScode agent for Airflow and n8n agent for Trino, helping in auto resolution of failed jobs. - Designing and deployed ComfyUI & InvokeAI instances under GenAI gateway ( UI – Next.js & backend FAST API) so that customers can auto provision instance in Kubernetes pod by triggering gitlab pipelines. - Designed and developed secure internal meta-registry and gateway (UI- Next.js & backend - LiteLLM) for MCP servers using FastMCP, SSO and middleware. - Designed and implemented algorithm to migrate non GPU workloads on Kubernetes to memory efficient CPU, thus bringing down costs across AWS EKS clusters for customer teams. - Designed and developed gitlab pipeline and for provisioning Label Studio and AIM open source tools on Kubernetes pods with Oauth reverse proxy as security mechanism. - Designed and implemented DARTS, a tool that automatically classifies Airflow DAG failures & Databricks job failures via log collection into buckets using DBSCAN & applies/select/creates appropriate rule from Rule classification engine to retry the error with updated infrastructure configuration. Thus, reducing dependency on L1/L2 support teams. - Designed and implemented product metric ETLs (Airflow DAGs) and Trino SCD tables to understand AI Platforms product acceptance within EA games by aggregating data from Prometheus, Loki, Databricks System Tables, Grafana, and GCP BigQuery. - Designed and implemented cost-effective chargeback solutions for in-house AI platforms and third-party products like Databricks, leveraging Python, PySpark, Jupyter Notebook, BigQuery, Airflow, and AWS, resulting in $3 million in savings for the AI platform team.

Software Engineer 1 | Backend at Demandbase, Mar 2020 - Sep 2022 · 2 yrs 7 mos, India, Working in this fast-paced startup environment is making my learning curve increase exponentially. Some of the things I have been involved in are: - Designed & implemented the Low Level Design for every user story/feature enhancement on API Product and underlying Core microservices. While also handling Sprint responsibilities for the team. - Building a microservice from scratch using Spring Boot, that runs Cron Jobs to populate the Redis Cache from SQL DB and writing JPA, Rest API's to expose data endpoints. - Writing automation suites/Unit tests using TestNg/Junit, and security bug proof code using Checkmarx with every development task to follow the agile methodology. - Adding new API endpoints to the existing API Product and debugging the issues. - Chalking out and implementing HLD for multiple microservices now, starting with upgrading vulnerable jars and Elasticsearch from 5.6.7 to 7 version and making required code changes. - Working rotationally with the team on Customer Escalations and providing engineering solutions to possible bugs or customer requests.

Software Engineer 1 at AIS (Applied Information Sciences), May 2019 - Feb 2020 · 10 mos, Hyderabad, Telangana, India

Education

Master of Technology - MTech, Artificial Intelligence, Birla Institute of Technology and Science, Pilani

Bachelor of Technology - BTech, Computer Science, UPES

Public links

github: https://github.com/Vinayak-Bajoria

medium: https://medium.com/@bajoriavinayak

bajoriavinayak.com: https://bajoriavinayak.com/#work