B.E. Computer Engineering, Fr. Conceicao Rodrigues College of Engineering (Fr. CRCE), Bandra, Mumbai, 2026
Experience
AI Development Intern · Siemens Digital Industries Software
Chennai · Feb 2026 – Jul 2026
Built an AI copilot that replaced menu-driven workflows with a conversational interface for complex version-control operations.
The conversational copilot (MCP). A conversational interface that replaced those menus, exposed both to the application UI and to external AI-enabled IDEs. An MCP server over a Spring Boot/gRPC backend that auto-generates LLM-discoverable tools from OpenAPI specs.
Tool discovery at scale. Hybrid intent routing plus semantic tool filtering that narrows 100+ candidate tools to 3–8 per request. A 384-dimension sentence-transformer embeds requests and tools; cut LLM API costs by ~75% while reducing inference load.
AI defect investigation. An automated pipeline that fetches Jira issues, resolves ambiguous component mappings and orchestrates AI-assisted root-cause analysis. It produces structured triage reports (suspect commits, impacted files, fix recommendations, test strategies) and auto-publishes them back to Jira. Now in regular team use. GitHub Copilot and Siemens-hosted LLMs, orchestrated in Python, so developers start from a pre-investigated issue.
Production engineering. A 3-stage GitLab CI/CD pipeline (lint → test → build) with packaged deployment. Automates 1,000+ test executions on every run.
100+ tools → 3–8 per request · ~75% lower LLM API cost · 8 repos · 1,000+ automated tests
Research Intern · Fr. CRCE Research
Bandra, Mumbai · Jun–Aug 2025 · Under Prof. Joshua Michel & Dr. Sujata Deshmukh
Led an evaluation of LLM fine-tuning strategies (LoRA, DPO, ORPO) across GPT-2, Gemma 3 (4B) and Llama 3.1 (8B); reached 78.6% accuracy with Llama 3.1 + DPO, 22% above baseline.
Technical Lead Intern · Modern Hub Solutions
Remote · Dec 2024–May 2025 · Awards: Best Rising Leader and Best Rising Star
Guided a 14-person team to ship a clinical platform with a RAG-based diagnostic chatbot on containerised microservices for real-time physician queries.
Data Engineer Intern · Jio Platforms
Navi Mumbai · Jun–Jul 2024
Device analytics with Apache Spark (Scala), Hive and MapReduce.
Recommendations
“When Shaun joined the development team for which I'm a product owner, I quickly realized the team hit the jackpot.”
“My biggest regret is that we were not able to hire him.”
“A motivated and quick learner who demonstrated strong curiosity, dedication, and a willingness to take on new challenges.”
Build a multilingual agent once from your docs, website and spreadsheets, then deploy it to browser voice, phone, WhatsApp, Telegram, a website widget, a hosted chat page, email and the API, with one customer memory spanning every channel.
One agent deployed to browser voice, phone, WhatsApp, Telegram, a website widget, a hosted chat page, email and the API.
One customer memory spanning every channel.
Built-in CRM, campaigns, automations, human handoff, analytics, evaluations and a glass-box trace of every answer.
Voice in 25+ languages with per-turn language detection (including Hinglish).
Evals covering Ragas-style QA, a 12-attack red team, simulated customers, and voice simulations scoring WER and p95 latency. CI fails below 70%.
Earlier version: Rebuilt and expanded from VoiceFlow AI: a multi-tenant voice SaaS with per-tenant RAG on ChromaDB + BM25, real-time Twilio/WebSocket streaming, FastAPI, LangGraph, Redis, PostgreSQL and MinIO.
A friendly AI money guide for India: ask anything about saving, investing, tax or retirement by voice or text, in any of 11 Indian languages, and get answers grounded in what Indian regulators actually publish.
A chat-and-voice money assistant on web and mobile (Expo / React Native).
Help with portfolio analysis, retirement and FIRE planning, tax optimisation, a financial-health score, life-event stress tests and regulatory questions.
Answers backed by RBI, SEBI and IRDAI material, retrieved and cited rather than guessed.
One assistant on the surface; 22 specialised agents behind it.
A 4-tier intent classifier (regex → embeddings → LLM fallback) that cut LLM API costs by 50%.
Kumbh Watch · Real-time crowd monitoring for mass gatherings.
Turns live camera feeds into a crowd-density map for gatherings the size of the Kumbh Mela, raises an alert when a zone behaves abnormally, and gives people an AI voice line for emergencies.
YOLOv8 person detection, density heatmaps and anomaly detection.
Live RTSP camera feeds over WebSockets.
An AI emergency voice service with a knowledge base.
🥇 Winner, SUNHACKS 2025
YOLOv8 · OpenCV · WebSockets · RTSP · Voice AI
TicketFlow AI · The first stage of IT ticket resolution, automated.
An AI help-desk assistant that reads every new IT ticket, files it under the right category, ranks how urgent it is and drafts a likely fix before a person opens it.
A four-step pipeline, each step with its own specialist: classify the ticket, set its priority, retrieve a likely solution, then review that answer's quality before anyone sees it.
Estimates each ticket's priority and how long it will take to resolve.
Not tied to one model provider: runs on Groq, Gemini, Ollama or Hugging Face.
LangGraph · Groq · Gemini · Ollama · Hugging Face
More departures
ResQ AI: A co-pilot for 112 call-takers: it listens to a distress call, judges how urgent it is and helps decide what to send where, while the operator watches it all on a live dashboard. Distress-call analysis with an urgency assessment for every call. Resource-allocation suggestions for the operator. A real-time operator dashboard. (🥇 Winner, Neural Nexus AI/ML Hackathon)
PagePulse: See how people really use your site (where they click, where they drop off, what went wrong) without sending their data to a third party. Heatmaps, funnels, cohorts, A/B tests, error tracking, PII masking and consent management. A custom browser tracking SDK.
Beacon GTM Lakehouse: A complete go-to-market data platform, from raw sales events to a live dashboard, that spins up on a laptop. Bronze → Silver → Gold medallion architecture in PySpark. Postgres serving layer, Redpanda streaming, Airflow orchestration and a Streamlit dashboard. All running in Docker with no cloud account needed.
rag-lab: A hands-on teaching platform that takes you from your first retrieval pipeline to production concerns like evaluation and security, one lesson at a time. 11 progressive lessons, from basic RAG to security, each a self-contained FastAPI module. Real retrieval in every lesson, with an extractive fallback so it runs without any API key. Optional extras: Docling, Crawl4AI, RAGAS and DeepEval.
CREDA: an AI financial-literacy advisor in 11 Indian languages
BE All-Rounder Award (Fr. CRCE, Jan 2026): Selected from the entire fourth-year BE batch for overall contribution across technical work, extracurriculars, communication and leadership.
Google AI for Impact (Asia-Pacific): Ranked Top 90 globally among 10,000+ participants.
IET Conference Proceedings, International Conference ICATES 2025 · Indexed in IEEE Xplore · DOI 10.1049/icp.2025.4777
Combines LSTM networks (behavioural patterns) with Graph Neural Networks (relational fraud across accounts), adaptive risk thresholding, and explainability via SHAP and LIME, with a mobile validation prototype and a Streamlit dashboard. First author.
Grenze International Journal of Engineering & Technology (GIJET), Vol. 11, Issue 1, 2025 · Scopus-indexed
An academic project platform with cosine-similarity plagiarism detection, submissions, reviews, ratings and student–faculty feedback. Co-author.
Community
GDSC Fr. CRCE (Google Developer Student Clubs, Fr. CRCE): joined as a web developer and became Vice Chairperson & Co-Lead. Co-led a 50-member team. Organised an international hackathon with participants from 20+ countries. Ran 10+ workshops on Web, Google Cloud and AI for 100+ students.
Mentor and judge: Bit N Build: GDG On Campus Fr. CRCE · Oct 2026
Hackathon mentoring: Helping juniors build their hackathon projects.
RAG: Beyond Chat with PDFs: A technical session for students.