Headline

Senior Backend + GenAI Engineer | FastAPI · RAG · Agentic AI · LangGraph | Qualcomm · Experian | Production APIs serving 2M+ users

About

Most engineers are either solid on backend or excited about AI. I do both and I've learned that the line between them barely exists when you're building systems that need to stay up. Over 5 years across Qualcomm and Experian, I've architected production systems that move real numbers: a RAG chatbot serving 2M+ users with 92% satisfaction, an agentic log-analysis platform that cut debug turnaround by 60% for 300+ engineers globally, an ASR pipeline that halved transcription costs ($80K → $40K/yr), and a GenAI email assistant that tripled SDR reply rates across 500+ B2B sales teams. The backend work is what keeps the AI work standing. 𝗕𝗮𝗰𝗸𝗲𝗻𝗱 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 → FastAPI microservices handling 1,000+ concurrent requests with async I/O, JWT auth, and structured error handling → Cloud infrastructure on AWS: SageMaker (20+ endpoints), Lambda, S3, with MLflow monitoring and 35% latency reduction → Databases across the stack: PostgreSQL, MongoDB, Redis, CosmosDB — chosen for the workload, not the default 𝗚𝗲𝗻𝗔𝗜 & 𝗟𝗟𝗠 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 → Production RAG architecture: GPT-4 + LangChain + CosmosDB vector search, chunking strategy, embedding pipeline, retrieval tuning — deployed at Experian at 2M+ user scale → Multi-agent orchestration: LangChain + CrewAI + Model Context Protocol (MCP) skill framework enabling 50+ engineers to ship GenAI features 3× faster → LLM fine-tuning: supervised fine-tuning on domain-specific B2B sales data with prompt engineering and structured output design → Edge AI: SNPE HTP operator implementation on Snapdragon NPUs — automated quantization debugger saving 50+ engineering hours per week 𝗪𝗵𝗮𝘁 𝘀𝗲𝘁𝘀 𝗺𝗲 𝗮𝗽𝗮𝗿𝘁 I've worked at every layer of the stack — from writing C++ operators on Snapdragon silicon to deploying GPT-4 pipelines to millions of users. That full-stack depth means I don't just build the AI feature; I own the infrastructure that makes it reliable, observable, and cost-efficient in production. I treat every inference endpoint as a backend API first — with rate limiting, retry logic, monitoring, and SLO accountability. 𝗘𝗱𝘂𝗰𝗮𝘁𝗶𝗼𝗻 M.E. in Computer Science — Jadavpur University (GPA: 8.83) B.E. in Computer Science — GCETT Berhampore (GPA: 9.02) I'm actively exploring Senior Backend Engineer, Senior GenAI Engineer, and AI Engineer roles at product-first companies — especially those scaling LLM-powered systems, agentic infrastructure, or distributed backend platforms. If your team is building something in that space, I'd love to connect. 📩 mondalayan96ju@gmail.com

Location

West Bengal

AI stack

codex

claudecode

Work experience

Senior Software Engineer at Qualcomm, Apr 2024 - Present · 2 yrs 5 mos, Hyderabad, - Architected an 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗹𝗼𝗴-𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝘀𝘆𝘀𝘁𝗲𝗺 using LangChain and a proprietary LLM to auto-diagnose CI build failures and recommend resolutions, 𝗰𝘂𝘁𝘁𝗶𝗻𝗴 𝗱𝗲𝗯𝘂𝗴 𝘁𝘂𝗿𝗻𝗮𝗿𝗼𝘂𝗻𝗱 𝗯𝘆 𝟲𝟬% for 300+ engineers globally. - Built an 𝗠𝗖𝗣-𝗽𝗼𝘄𝗲𝗿𝗲𝗱 𝘀𝗲𝗹𝗳-𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗮𝗴𝗲𝗻𝘁 on top of Release Confluence and MongoDB, enabling natural-language queries over internal release data — 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗰𝗿𝗼𝘀𝘀-𝘁𝗲𝗮𝗺 𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝘆 𝗯𝘆 𝟮𝟱% and eliminating ad-hoc data requests across engineering orgs. - Migrated 𝟭𝟱𝟬+ 𝗳𝗿𝗲𝗲𝘀𝘁𝘆𝗹𝗲 𝗝𝗲𝗻𝗸𝗶𝗻𝘀 𝗷𝗼𝗯𝘀 to declarative pipelines with standardised templates, 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁 𝗲𝗿𝗿𝗼𝗿𝘀 𝗯𝘆 𝟰𝟬% and establishing CI/CD standards adopted org-wide. - Designed an 𝗠𝗖𝗣 𝘀𝗸𝗶𝗹𝗹 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 enabling 50+ engineers to ship GenAI features 𝟯× 𝗳𝗮𝘀𝘁𝗲𝗿 using reusable LLM-backed tooling — cutting average feature delivery from weeks to days.

Software Engineer at Experian, Dec 2023 - Apr 2024 · 5 mos, Hyderabad, - Architected and deployed a 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗥𝗔𝗚 𝗰𝗵𝗮𝘁𝗯𝗼𝘁 (GPT-4 + LangChain + CosmosDB vector search) explaining credit-score changes in plain language — shipped to 𝟮𝗠+ 𝘂𝘀𝗲𝗿𝘀 𝗶𝗻 𝘂𝗻𝗱𝗲𝗿 𝟲 𝘄𝗲𝗲𝗸𝘀 with 𝟵𝟮% 𝘂𝘀𝗲𝗿 𝘀𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻 at launch. - Standardised 𝟮𝟬+ 𝗺𝗼𝗱𝗲𝗹 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁 𝗲𝗻𝗱𝗽𝗼𝗶𝗻𝘁𝘀 via AWS SageMaker with MLflow monitoring, 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗶𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗹𝗮𝘁𝗲𝗻𝗰𝘆 𝗯𝘆 𝟯𝟱% 𝗮𝗻𝗱 𝗟𝗟𝗠 𝗔𝗣𝗜 𝗰𝗼𝘀𝘁𝘀 𝗯𝘆 𝟯𝟬% across all production pipelines.

Data Scientist at Outplay, Jun 2023 - Aug 2023 · 3 mos, Hyderabad, - Deployed an 𝗔𝗜 𝗘𝗺𝗮𝗶𝗹 𝗔𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁 (fine-tuned GPT + RAG pipeline) across 500+ B2B sales teams, 𝘁𝗿𝗶𝗽𝗹𝗶𝗻𝗴 𝗦𝗗𝗥 𝗿𝗲𝗽𝗹𝘆 𝗿𝗮𝘁𝗲𝘀 𝗳𝗿𝗼𝗺 𝟭𝟱% → 𝟰𝟱% — directly accelerating revenue pipeline for every customer account. - Built an 𝗶𝗻-𝗵𝗼𝘂𝘀𝗲 𝗔𝗦𝗥 𝘀𝘆𝘀𝘁𝗲𝗺 𝘄𝗶𝘁𝗵 𝘀𝗽𝗲𝗮𝗸𝗲𝗿 𝗱𝗶𝗮𝗿𝗶𝘀𝗮𝘁𝗶𝗼𝗻 from scratch in approximately 1 month, replacing a third-party vendor and 𝗰𝘂𝘁𝘁𝗶𝗻𝗴 𝘁𝗿𝗮𝗻𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 𝗰𝗼𝘀𝘁𝘀 𝗯𝘆 𝟱𝟬% ($80K → $40K/yr) at 95% accuracy. - Developed 𝗕𝗘𝗥𝗧-𝗯𝗮𝘀𝗲𝗱 𝘁𝗼𝗻𝗲-𝗮𝗱𝗷𝘂𝘀𝘁𝗺𝗲𝗻𝘁 𝗺𝗼𝗱𝗲𝗹𝘀 with an A/B testing framework, 𝗶𝗺𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗲𝗺𝗮𝗶𝗹 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 𝗿𝗮𝘁𝗲𝘀 𝗯𝘆 𝟮𝟬% across live customer cohorts.

Associate Data Scientist at Outplay, Jul 2022 - May 2023 · 11 mos, Hyderabad, Telangana, India

AI/ML Engineer at Qualcomm, Jul 2021 - Jul 2022 · 1 yr 1 mo, Hyderabad, - Enabled 𝟯𝟬+ 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗠𝗟 𝗺𝗼𝗱𝗲𝗹𝘀 on Snapdragon devices by implementing ONEHOT, LSTM, and Detection Out operators in 𝗦𝗡𝗣𝗘 𝗛𝗧𝗣 — directly unblocking 𝗺𝗼𝗯𝗶𝗹𝗲 𝗲𝗱𝗴𝗲-𝗔𝗜 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗶𝗮𝗹𝗶𝘀𝗮𝘁𝗶𝗼𝗻 across multiple product lines. - Built an 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗾𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗱𝗲𝗯𝘂𝗴𝗴𝗶𝗻𝗴 𝘁𝗼𝗼𝗹 eliminating manual trial-and-error across SNPE configurations, 𝘀𝗮𝘃𝗶𝗻𝗴 𝟱𝟬+ 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗵𝗼𝘂𝗿𝘀 𝗽𝗲𝗿 𝘄𝗲𝗲𝗸 for the entire SNPE team. - Delivered an 𝗡𝗟𝗣-𝗽𝗼𝘄𝗲𝗿𝗲𝗱 𝗿𝗲𝘀𝘂𝗺𝗲 𝗽𝗮𝗿𝘀𝗲𝗿 (BERT + spaCy + HuggingFace) adopted as the 𝗼𝗳𝗳𝗶𝗰𝗶𝗮𝗹 𝗿𝗲𝗰𝗿𝘂𝗶𝘁𝗺𝗲𝗻𝘁 𝘁𝗼𝗼𝗹 𝗮𝗰𝗿𝗼𝘀𝘀 𝟭𝟬+ 𝘁𝗲𝗮𝗺𝘀, screening 100+ candidates per month and saving 10+ HR hours per week organisation-wide. - Engineered an 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗝𝗜𝗥𝗔 𝗵𝗮𝗻𝗱𝗹𝗲𝗿 using BERT-based NLP for intelligent issue routing, 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗺𝗮𝗻𝘂𝗮𝗹 𝗶𝘀𝘀𝘂𝗲-𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗲𝗳𝗳𝗼𝗿𝘁 𝗯𝘆 𝟳𝟬% and accelerating triage velocity across engineering orgs.

Data Scientist at BarRaiser, Mar 2021 - Jun 2021 · 4 mos, Bangalore Urban, - Solely architected and delivered 𝟯 𝗶𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗔𝗜 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 within a 4-month internship — video analysis, anomaly detection, and tag suggestion — each deployed on AWS end-to-end. - Built an 𝗔𝗪𝗦 𝗘𝗹𝗮𝘀𝘁𝗶𝗰 𝗕𝗲𝗮𝗻𝘀𝘁𝗮𝗹𝗸 𝘃𝗶𝗱𝗲𝗼-𝗶𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 processing Zoom recordings end-to-end, analysing 𝟱𝟬𝟬+ 𝗶𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄𝘀 𝗽𝗲𝗿 𝗺𝗼𝗻𝘁𝗵 with 90% question-extraction accuracy — eliminating manual QA review entirely. - Architected 𝗙𝗮𝘀𝘁𝗔𝗣𝗜 𝗶𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗺𝗶𝗰𝗿𝗼𝘀𝗲𝗿𝘃𝗶𝗰𝗲𝘀 supporting 𝟭,𝟬𝟬𝟬+ 𝗰𝗼𝗻𝗰𝘂𝗿𝗿𝗲𝗻𝘁 𝗿𝗲𝗾𝘂𝗲𝘀𝘁𝘀 with sub-second response times providing the production backbone for real-time interview analysis at scale. - Engineered an 𝗮𝗻𝗼𝗺𝗮𝗹𝘆-𝗱𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺 identifying score-comment mismatches in interview feedback data, 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗿𝗲𝘃𝗶𝗲𝘄𝗲𝗿 𝗲𝗿𝗿𝗼𝗿 𝗯𝘆 𝟯𝟱% and cutting QA analysis time across all interviewer performance records. - Developed a 𝗕𝗘𝗥𝗧-𝗯𝗮𝘀𝗲𝗱 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻-𝘁𝗮𝗴 𝘀𝘂𝗴𝗴𝗲𝘀𝘁𝗶𝗼𝗻 𝗲𝗻𝗴𝗶𝗻𝗲, 𝗿𝗲𝗱𝘂𝗰𝗶𝗻𝗴 𝗺𝗮𝗻𝘂𝗮𝗹 𝘁𝗮𝗴𝗴𝗶𝗻𝗴 𝗲𝗳𝗳𝗼𝗿𝘁 𝗯𝘆 𝟲𝟬% and improving insight accuracy across the interviewer scoring pipeline.

Education

Master of Technology - MTech, computer science and engineering, Jadavpur University

Bachelor of Technology - BTech, Computer Engineering, Government College of Engineering and Textile Technology, Berhampore