Portfolio - Gaurav Singh

Professional Summary

AI Platform Engineer building production distributed systems on Temporal, Argo Workflows, and Java at Atlan — cut workflow MTTR by 47% via AI-native observability pipelines and reduced customer ticket TTR from 30 hours to 5 hours through an agentic Claude Code SDK orchestration system. GSoC 2025 contributor (approximately 3% acceptance rate) who shipped a multi-LLM code-review platform with 300%+ response-time improvement across 10+ benchmarked models. Specializes in self-healing workflow architecture, P99 latency profiling, RAG-powered semantic search (sub-150ms over 100K+ embeddings), and agentic AI systems embedded into engineering and support pipelines.

Skills

Languages & Frameworks: Golang, Python, Java, JavaScript, Dart (Flutter), LangChain, OpenAI SDK, NextJS, NodeJS

AI / ML: Generative AI, RAG, AI Agents, NLP, LLM Fine-tuning, Quantized Models (GGUF), Vector DBs (FAISS, Pinecone), Gemini API

Platform & Infra: Temporal, Argo Workflows, Kubernetes, Docker, GitHub Actions, AWS (DynamoDB, EC2), GCP, CI/CD

Databases & Observability: PostgreSQL, AWS DynamoDB, Vector DBs, JVM Profiling, Anomaly Detection, Distributed Tracing

Domain Expertise: Distributed Workflow Orchestration, AI-Native Observability, Backend Architecture, Self-Healing Systems, ETL Pipelines

Experience

Atlan (Remote)

AI Platform Engineer — App Runtime Team | Nov 2025 – Present

Engineered AI-native workflow analytics and observability pipelines to automatically triage complex query and workflow failures, cutting MTTR by 47% through proactive anomaly detection, alerting, and structured root-cause reports.
Built and maintained distributed workflow orchestration features leveraging Temporal, Argo Workflows, and Java-based services, improving reliability and horizontal scalability of large-scale data and analytics pipelines serving enterprise-scale tenants.
Developed a self-healing workflow mechanism with automated recovery actions on P99 latency breaches and vital metric violations, eliminating an entire class of on-call escalations and reducing incident volume by approximately 60%.
Engineered a Ticket-to-PR Agentic Orchestration System using a sandboxed context layer with the Claude Code SDK, autonomously triaging customer issues and opening fix PRs—slashing customer ticket TTR from 30 hours to 5 hours (an 83% reduction).
Partnered with internal and external customers to debug production incidents in distributed runtimes; fed learnings back into design documents and PRs, driving continuous platform improvement.

Google Summer of Code — Keploy (California, USA, Remote)

Open-Source Contributor / AI Engineer | Jun 2025 – Sep 2025

Spearheaded an open-source multi-LLM code-review platform with NLP-powered inline static analysis, improving developer productivity by 25% and earning a GSoC selection (competitive approximately 3% acceptance rate).
Architected a fully automated CI/CD pipeline with GitHub Actions runners, achieving 99.9% operational uptime and zero-touch code review on every PR.
Benchmarked 10+ open-source LLMs for accuracy, latency, and context fit; optimised response times by 300%+ via efficient chunking algorithms and GitHub-hosted model integrations.
Deployed quantized models in GGUF format to eliminate local overhead, reducing memory footprint by 70% and cutting response latency by 100ms.

Code for GovTech (C4GT’25) — OpenNyAI (Bengaluru, India)

AI Engineer | Jun 2025 – Sep 2025

Unified vector databases to store 100,000+ embeddings, powering real-time semantic search with sub-150ms latency across constitutional law documents.
Refined query accuracy by 70% through fine-tuning ranking algorithms and advanced intent recognition techniques.
Engineered an ETL pipeline processing 10,000+ records/hour with 85% noise reduction via a modular adapter-dispatcher protocol.
Authored an AI-powered platform to track 100+ Indian Constitution amendments, featuring NLP-based summarisation and version control.

Code for GovTech (C4GT’24) — eGov Foundation (Remote)

Software Intern | Jun 2024 – Nov 2024

Built a voice-enabled Flutter UI with NLP, reducing voice-input latency by 20% for a Digital Public Goods initiative targeting 2M+ global users.
Integrated speech-to-text and NLP libraries with backend APIs, sustaining a responsive Flutter UI for 10,000+ daily active users (DAUs).

Projects

Gollora | Golang, Gemini API, golangci-lint, staticcheck, gosec | GitHub

Designed an AI-engineered code-review agent for GitHub workflows that accelerated feedback delivery by 5× and lifted code quality metrics by 35% via merged static analysis tools.
Implemented Google Gemini API for smart code review, raising bug and security-issue detection accuracy by 40%; achieved 98% code coverage.

linkedin-nugget | LLM, Agentic AI, NLP | GitHub

Created an agentic AI LinkedIn post generator with 100% tone/formatting fidelity and data-driven hashtag selection, increasing post creation speed by 5× and expanding average reach by 40%.

ell-ena | Flutter/Dart, Gemini AI, NLP | GitHub

Constructed an AI-integrated product-manager assistant automating task creation via natural language (4× faster), with 99% task-type extraction accuracy and 60% reduction in meeting-transcription time.