I work at the intersection of backend engineering, machine learning infrastructure, and generative AI. My focus is on building systems that stay performant under real production traffic while remaining maintainable for the teams that operate them.
AI/ML Backend Engineer
Surbhi Suman builds fast, resilient systems for LLM and ML products.
3+ years of experience shipping high-performance Go and Python microservices, production-grade recommendation systems, and automated MLOps pipelines for large-scale platforms.
About
Engineering backend intelligence for real-world scale
Across recommendation systems, LLM-serving APIs, and autonomous chatbot flows, I’ve delivered measurable business impact with a strong emphasis on concurrency, reliability, and operational efficiency.
Experience
Recent roles and high-impact work
Lead Engineer
- Reduced API latency by 82% by deploying high-concurrency Go microservices on GKE for LLM-powered suggested replies.
- Cut seller response time by 70% by architecting an autonomous WhatsApp chatbot using Gemini-based LLM agents.
- Reduced Docker container memory usage by 90%, from 155 MB to 16 MB, by migrating logging from file-based storage to stdout.
GoPythonLLMFAISSDockerGKE
Senior Software Engineer
- Drove a 17% uplift in buyer click-through rate with an ML recommendation system across 4 crore+ products and 54 TB of clickstream data.
- Built a hybrid recommender agent using a TensorFlow Two-Tower model, Gemini 2.0 Flash, Google ADK, and FAISS for semantic retrieval.
- Accelerated retraining cycles by 40% with automated PySpark-based ML pipelines on AWS EMR.
TensorFlowPySparkAWS EMRGeminiGoogle ADK
Software Engineer
- Improved backend data retrieval latency by 25% by optimizing SQL datamarts and caching strategies across high-traffic APIs.
- Built Kafka-driven ingestion and synchronization flows to keep downstream systems refreshed with low-latency updates.
SQLKafkaCachingMicroservices
Software Developer Intern
- Automated AWS CDK-based alarming systems and reduced mean time to recovery by 30%.
- Enhanced CI/CD pipelines using Ruby and TypeScript for faster, more reliable deployment workflows.
AWS CDKRubyTypeScriptCI/CD
Projects
Selected systems and platform work
Infrastructure
High-Throughput LLM Inference Proxy
Architected a high-concurrency Go proxy layer that orchestrates requests across providers like Gemini and OpenAI, with circuit breakers and exponential backoff for 99.9% availability.
Platform
Multi-Tenant MCP Gateway
Designed a scalable MCP gateway with FastAPI, Redis, and PostgreSQL, enabling tenant isolation, token-based authentication, rate limiting, and audited LLM tool routing.
Skills
Tools, platforms, and engineering strengths
Languages
Python, Go, SQL, C, TypeScript, Ruby, JavaScript
Backend
Microservices, REST APIs, Flask, Kafka, gRPC, Git, MLflow
Cloud
AWS EMR, Athena, S3, GKE, Vertex AI, Gemini
Infrastructure
Docker, Kubernetes, Devtron, CI/CD Pipelines, MLOps
Core CS
Graphs, Trees, Dynamic Programming, Arrays, DSU, Hashmaps
Education
Academic foundation
Master of Technology in Computer Science
IIIT Lucknow
2022 · CGPA 8.94Bachelor of Technology in Computer Science
CV Raman College of Engineering
2020 · CGPA 8.61Contact