I build production AI systems that turn operational noise into reliable, explainable action.
multi-agent orchestration · hybrid RAG · AI automation · cloud architecture · full-stack product engineering
| Since 2022 production engineering |
≈20% faster operational turnaround |
≈50% lower SME dependency |
74 credentials live-verifiable mastery |
Enterprise AI becomes valuable only when it can retrieve the right evidence, route work to the right specialist, survive imperfect inputs, and produce an action someone can trust.
I work across that entire path:
enterprise signals → intent routing → specialist agents → retrieval + reasoning
→ guarded automation → observable business outcomes
Designed and shipped a production multi-agent operational intelligence platform spanning incident, problem, release, testing, and service-request workflows.
- Orchestrated specialist agents with LangGraph, dynamic intent routing, contextual memory, and semantic retrieval.
- Built RAG pipelines over Jira, Confluence, SharePoint, GitHub, Splunk, SQL systems, and enterprise databases—covering structured and unstructured evidence.
- Developed real-time root-cause analysis that correlates logs with historical incidents and generates context-grounded resolution steps.
- Automated functional test generation plus Selenium, UFT, and Karate scripts from design documents and existing frameworks.
- Generated release notes, deployment plans, and rollback strategies from commits, tickets, and documentation.
- Implemented chat-to-ServiceNow ticket creation, natural-language-to-SQL analysis, anomaly detection, and audit-ready SOX reporting.
- Delivered a unified conversational interface backed by MongoDB session memory—helping reduce turnaround time by ≈20% and SME dependency by ≈50%.
Open architecture inventory
- AI orchestration: LangGraph, LangChain, multi-agent state machines, prompt/tool routing
- Knowledge: hybrid retrieval, vector search, semantic indexing, structured + unstructured ingestion
- Data: MongoDB, PostgreSQL, Databricks, SQL generation and impact analysis
- Enterprise integrations: Jira, ServiceNow, Confluence, SharePoint, GitHub, Splunk
- Automation: Selenium, UFT, Karate, release intelligence, compliance reporting
- Backend: Python APIs, asynchronous pipelines, conversation state, observability patterns
- Built modular Spring Boot / Core Java microservices and REST APIs for insurance workflows.
- Designed event-driven components with AWS Lambda, reducing infrastructure overhead.
- Automated build, test, and deployment through CI/CD and AWS CodeBuild workflows.
- Increased reliability with JUnit unit/integration tests and early quality gates through SonarQube and SourceClear.
- Contributed automation scripts and iterative delivery practices in an Agile, DevOps-first engineering team.
|
Enterprise document intelligence with retrieval depth. Hybrid BM25 + FAISS search, reciprocal-rank fusion, cross-encoder re-ranking, contextual chunk enrichment, multi-query expansion, OCR failover, and WebSocket streaming.
|
Location-aware environmental intelligence for greener cities. Combines weather, air quality, soil, biodiversity, climate forecasting, visual plant diagnosis, and garden-planning tools.
|
|
Multi-format RAG workspace. Hybrid retrieval, adaptive search, conversation memory, runtime key switching, responsive themes, rich tables, and chat export.
|
Document reasoning with observability. OCR fallback, contextual windows, hybrid search, reflection, analytics, session isolation, caching, and circuit breakers.
|
More experiments and shipped products
- Handwritten Chemical Compound Detection — TensorFlow object detection for hand-drawn chemical structures.
- Anupam Roy Portfolio — recruiter-facing AI systems command center with a mastery wall and AI consultant.
- React + Next.js Portfolio — responsive personal portfolio.
- Motion + Sanity Portfolio — React, SCSS, Framer Motion, and Sanity.
- Real Estate — React and Next.js property experience.
- Modern GPT-3 UI — responsive React interface study.
| Layer | Working toolkit |
|---|---|
| Agentic AI & LLM systems | LangGraph · LangChain · RAG pipelines · Gemini API · vector search · prompt engineering |
| Backend & data | Python · FastAPI · Java · Spring Boot · PostgreSQL · MongoDB · Databricks · Node.js |
| Cloud & platform | GCP · Azure · AWS · cloud architecture · Docker · CI/CD · serverless systems |
| Product engineering | React · Next.js · TypeScript · Streamlit · Tailwind · Framer Motion |
| Quality & automation | JUnit · SonarQube · Selenium · UFT · Karate · GitHub workflows |
Enter the live verification vault →
Rather than turning the profile into a badge catalogue, the first screen shows only high-signal proof. The vault contains issuer-hosted verification links across AWS, Anthropic, Databricks, Google Cloud, Microsoft, MongoDB, and Snowflake.
Signature proof: AWS Machine Learning Engineer · Claude Certified Architect · Databricks Generative AI Engineer · Google Professional ML Engineer · Google Professional Cloud Architect · Azure AI Engineer · SnowPro Gen AI.
- Credly Top Badge Earner 2024 with a collection of 180+ digital badges — verify recognition · view Credly
- Google Cloud learning legend: all three Technical Expert badges and four consecutive learning seasons — milestone story
- CodeChef 5★, peak rating 2109 — competitive programming profile
B.Tech, Computer Science & Engineering · Maulana Abul Kalam Azad University of Technology · 9.14/10 · 2018–2022
Core study: data structures & algorithms, operating systems, DBMS, computer networks, machine learning, software engineering, and distributed systems.
I am most interested in problems where retrieval quality, agent orchestration, backend reliability, security, and product experience all matter at once.
Start a conversation · Explore the portfolio · Connect on LinkedIn
Designed as an evidence-first systems map—not a résumé pasted into Markdown.