Pulkit Kejriwal

Resume

Atlassian

Dec 2025 – Present

Bengaluru, India

Designed a two-stage agentic query engine for heterogeneous production databases: a first LLM call performs semantic table selection and generates a structured Intermediate Representation executed via a deterministic IIR→Cube transform — ensuring no raw SQL exposure — followed by a second LLM call that generates analysis code and selects visualizations to render insights. Validated by Head of Engineering; nominated for internal patent review.
Contributed to an AI developer productivity platform: integrated multiple AI tools (Codex, Claude, etc.) into the OTel collector layer, proposed a bloom filter to bypass string matching on session data transforms, built the hourly Airflow DAG for per-PR AI usage aggregation, and extended the GraphQL serving layer to expose developer AI usage telemetry to engineering leadership.
Built an automated incident triage pipeline: on-call engineer provides a Playwright script of the failing flow; the tool binary searches across 500 deployments by replaying the script to isolate the culprit deployment, then an LLM ranks the resulting 50 commits to surface 4 high-risk candidates — cutting triage time by 85%.
Engineered a diagnostic tool comparing healthy vs. degraded feature flag configurations across live instances, achieving 98% noise reduction on unprotected flag changes. Recognized as ShipIt India Winner.

Atlassian

July 2024 – Dec 2025

Bengaluru, India

Analysed 100+ alerts and redesigned org-wide monitoring using sliding window detection and incremental alerting to deduplicate redundant signals; cut Time-to-Detect from 4+ hours to under 30 minutes, raised detection probability from 10% to 33%, and eliminated 70% of noisy alerts.
Led AI-powered field pre-filling for Jira issue creation, enabling any AI agent to interact with the Jira issue creation modal; worked cross-functionally with Product, Design, and third-party consumers to migrate enterprise customers to the modern stack.
Served as feature lead for the Teamwork Collection preview panel — a flagship initiative unifying first- and third-party Atlassian products into a single surface across 100M+ users.

Atlassian

July 2023 – July 2024

Bengaluru, India

Migrated Jira's issue creation and transition interfaces from JSP to React and GraphQL, improving performance by 57% and unblocking enterprise migrations to Jira Cloud; developed GraphQL field data fetchers and reusable components powering features across 22M+ daily users via 10+ integrations.
Mentored 3 winter interns across technical and professional development; all three received pre-placement offers.

B.Tech in Computer Science and Engineering · CGPA: 8.58

July 2019 – May 2023

Ranchi, India

LanguagesC++, JavaScript (ES6+), Kotlin, Python, Go

AI & MLLLM-Agnostic Architecture, Retrieval-Augmented Generation, Agentic Pipelines

Data & StreamingKafka, Flink (Stateful Streams), Airflow, Databricks, StarRocks, Amplitude

Backend & SystemsDistributed Systems, GraphQL/Relay, REST APIs, OpenTelemetry, Terraform

Frontend & ObservabilityReact, Datadog/SignalFx, Feature Flag Systems