Apurv Garg

Software Engineer

Gurugram, Haryana, India4 yrs 1 mo experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Shipped agentic RAG system reducing manual review by 80%.
Enhanced retrieval quality with advanced tuning techniques.
Prototyped RAG stack improving internal demo capabilities.

Stackforce AI infers this person is a SaaS expert specializing in AI-driven retrieval systems and data processing.

Contact

Skills

Core Skills

Retrieval-augmented Generation (rag)Azure OpenaiPython

Other Skills

AngularAngular 19Angular MaterialApache KafkaArtificial Intelligence (AI)CC (Programming Language)Cascading Style Sheets (CSS)CybersecurityData AnalysisData StructuresDatabase Management System (DBMS)DockerElasticsearchGrafana

About

I build prompt-first RAG and reliable tool-calling systems. I design system prompts and curated few-shots, enforce schema-validated JSON + guardrails, and tune PGVector retrieval (chunking, filters, re-ranking) for grounded, cited answers. I automate Python/SQL batch runs on versioned golden sets (with targeted synthetic cases) and track precision, recall, F1, recall@k, latency, and cost so improvements ship fast without regressions. Recently: shipped agentic RAG in MDO (Azure OpenAI + PGVector + Bing-grounded search + Playwright evidence) with URL/screenshot citations — cut manual review effort by 80% and lifted grounded-answer rate by 80%. Open to roles where prompt engineering, retrieval quality, and practical evaluation matter.

Experience

Prospecta software

2 roles

Senior Development Engineer

Promoted

Mar 2025 – Present · 1 yr · Gurugram, Haryana, India

Shipped an agentic RAG system for MDO. Integrated Azure OpenAI chat and embeddings with PGVector retrieval, Bing-grounded web search, and Playwright evidence capture. Added URL and screenshot citations in the UI. This reduced manual review effort by 80% and improved grounded answer rate by 80%.
Designed system prompts and curated few-shot sets. Enforced strict JSON schemas and guardrails such as cite-before-answer and safe refusals.
Implemented reliable tool and function calling with typed inputs and outputs, retries, and clear error paths across search, classification, and normalization flows.
Tuned retrieval with better chunking, metadata filters, and re-ranking. Validated improvements on versioned golden sets using batch runs.
Built a Python and SQL batch runner and CLI with dataclasses, tests, and structured logging. CI publishes HTML diffs for every change.
Tracked precision, recall, F1, recall at k, p90 latency, and cost. Used these metrics to accept or reject changes.

Azure OpenAIPGVectorPythonSQLJSONRetrieval-Augmented Generation (RAG)+1

Devlopment Engineer

Feb 2022 – Mar 2025 · 3 yrs 1 mo · Gurugram, Haryana, India

Prototyped the RAG stack on Postgres with PGVector. Built Python embedding and ingestion jobs with Azure OpenAI. Created a first-pass retriever using chunking and metadata filters for internal demos.
Created the first golden set with about two hundred labeled cases. Wrote notebooks to compute precision, recall, F1, and recall at k. Set up a nightly CI job that shared HTML diffs.
Drafted early system prompts and few-shot libraries for classification and description. Enforced JSON schema outputs and a clear cite-before-answer rule.
Added Playwright based evidence capture to fetch supplier pages and PDFs. Stored URL and screenshot citations with each prediction.
Upgraded Java from 11 to 21 and Spring Boot from 2 to 3. Improved API throughput and efficiency.
Scaled Kafka by redesigning topics and consumer groups. Handled millions of daily events without downtime.
Tuned Elasticsearch aggregations and indexes. Reduced heavy report latency from minutes to seconds.
Modernized the Angular app with server side rendering and route level code splitting. Improved first paint and overall responsiveness.
Introduced Redis caching and Prometheus and Grafana dashboards. Maintained CI and CD with Jenkins and GitHub Actions.

Azure OpenAIPGVectorPythonPostgreSQLJavaKafka+3

Cisco

Cisco Cybersecurity

Apr 2021 – Jun 2021 · 2 mos · Ghaziabad, Uttar Pradesh, India

Gained hands-on exposure to modern cyber security frameworks, vulnerability scanning, and risk assessment protocols.
Assisted in identifying and mitigating high-priority vulnerabilities in internal tools.
Worked with cross-functional security and product teams to implement multi-layered security measures ensuring compliance with industry standards.

CybersecurityVulnerability ScanningRisk Assessment