Sai Nageswar Satchidanand

AI Researcher

Hyderabad, Telangana, India13 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in Deep Learning and ML-Pipelines.
Proven track record in building scalable systems.
Strong background in Kubernetes and distributed systems.

Stackforce AI infers this person is a Deep Learning and Full-Stack Developer with expertise in SaaS and E-commerce.

Contact

sainageswar@gmail.com +917022374478 LinkedIn

Skills

Core Skills

Deep LearningMl-pipelinesKubernetesDistributed SystemsCausality AnalysisSource Code AnalysisFull-stack DevelopmentMiddleware DevelopmentApi Development

Other Skills

AngularApache SparkCI/CDData StructuresDevOpsFunctional TestingGitGo (Programming Language)JavaKotlinModel DevelopmentObservabilityPipelinesPythonREST APIs

About

A passionate developer with experience in Deep learning, Big data & Full-Stack. Well experienced in System Design, building ML-Pipelines, gRpc APIs, Devops (Infra-As-Code), Kubernetes, Queues/Topics, Sql/NoSql Databases, Operational Excellence etc. In Deep Learning worked on variety of problems like Causality Analysis, Images, Source Code etc. and engineered distributed system for Feature Engineering and Model Inference at scale.

Experience

Microsoft

2 roles

Senior Applied Scientist

Promoted

May 2022 – Present · 3 yrs 10 mos

1. Developed offline pipelines for generating LLM-based titles, captions, and web answers for Bing's top search queries. Engineered a vector index using HNSW to cache this content and serve it on the Search Results Page via nearest neighbor searches. Additionally, experimented with various embedding queries to optimize performance.
2. Built online LLM captions generation without impacting overall page rendering time. Built a high scale system that triggers LLM preemptively before ranking and save inferences to a distributed cache. As ranking is finished, generated captions would be picked up from cache. Built comprehensive observability to record metrics for each step to optimize system end-to-end latency.
3. Enhanced Bing Search Results by using LLMs by chaining prompts for intent generation, alternative queries, and re-ranking. This utilized Chain of Thought Prompting, where the LLM was guided step-by-step to provide results with reasoning. Additionally, developed a streaming engine that sends LLM output tokens to the UI, masking the reasoning.

Deep LearningML-PipelinesgRpc APIsDevOpsKubernetesObservability

Senior Software Engineer

Jan 2021 – May 2022 · 1 yr 4 mos

Built Kubernetes for Kubernetes.
As Kubernetes mgmt plane handles resources like pods, deployments, services in multiple nodes, our system handled multiple Kubernetes clusters across multi-location, multi-tenant and multiple types (GPU clusters, Memory Optimized Clusters etc.)
Similar to Kubernetes, our distributed system handled reconciling actual state of multiple clusters with their desired state.
Further it handled resources like Keyvault, common containers in pods for logging, monitoring, Api-to-Api auth etc.
Our system had spec definition framework similar to Kubernetes, however, typesafe and general programming language based instead of yaml/json.

KubernetesDistributed SystemsResource Management

Amazon

SDE II - ML

May 2017 – Dec 2020 · 3 yrs 7 mos · Bangalore

Worked in Causality Analysis using Deep Learning. Our team was entrusted to measure effectiveness of ads posted in Search Engines, Social Sites by Amazon. To achieve this, we identified Control Variables (intrinsic features of customers like age, recency, frequency, years on platform etc.) and Treatment Variables (ads seen/clicked by Customer). Learned a model to map these variables with outcome – user action. Then measured incremental affect of treatment variables on outcome conditional to control variables. Also, employed several techniques to keep explain-ability of model consistent across re-training by developing custom cost functions that regularize weights with previously trained model.
Worked both in model development and developing pipelines for model inferencing at scale using SWF, ECS, AWS-Batch, EMR-Spark, Tensorflow etc.

Deep LearningCausality AnalysisModel DevelopmentPipelines

Codenation

2 roles

SDE II

Promoted

May 2016 – May 2017 · 1 yr · Bengaluru Area, India

Worked in Source Code Analysis. To apply deep learning on source code, developed embedding of source code on AST similar to word embedding in NLP.
Besides this worked on developing automated system for measuring multiple source code quality metrics. Used best of the development practices along with devops using Docker, ECS, etc.

Deep LearningSource Code AnalysisDevOps

SDE

Jun 2014 – Jul 2015 · 1 yr 1 mo · Dubai

1. Worked on developing an Airport app that helps travellers to store ticket, assist with navigation in airport till boarding of aircraft.
2. Worked on an enterprise application that gives statistical analysis of accounting data. Was mostly responsible for middleware and was involved in migration of legacy codebase - Upgrading Tech Stack, Adding tests, Adding CI/CD etc.

Middleware DevelopmentCI/CDStatistical Analysis

Zoomo (gozoomo.com)

Developer

Jul 2015 – May 2016 · 10 mos · Bengaluru Area, India

Worked as a Full Stack Dev developing a peer-to-peer trustworthy marketplace for used cars. Worked on mobile application to inspect cars , backend REST API microservices, frontend angular responsive website, crawler and operational excellence.

Full-Stack DevelopmentREST APIsAngular

Infosys

Systems Engineer

Jul 2010 – Jul 2012 · 2 yrs · Mysuru Area, India

Worked for Apple in iCloud core services. Developed APIs to reliably store files with de-duping. Worked on high quality code with best of the practices involving functional testing, CI/CD, Integration testing etc.

API DevelopmentFunctional TestingCI/CD