> whoami
I build intelligent systems, ship fast products, and sometimes make computers do things they weren't supposed to. RAG fan. Open-source contributor. Perpetually caffeinated.
01. about_me
Results-oriented Data Scientist & AI Engineer with 8+ years of experience tackling complex business problems across diverse industries. Expertise in advanced analytics, machine learning, statistical modelling, and LLM-powered systems.
Currently building RAG architectures and LLM evaluation tools as a freelance consultant, working with vector databases like ChromaDB, Pinecone, and Weaviate. Previously drove measurable impact at AB InBev (12.6% revenue uplift) and Piramal Glass ($500k cost savings).
MSc in AI & ML from LJMU Liverpool. Certified in Azure DP-100, AWS Solutions Architect, and Google Analytics.
$ cat stats.json
"experience": "8+ years",
"companies": 4,
"certifications": 4,
"coffee_consumed": "∞"
02. skills
ls skills/
03. timeline
Apr 2024 – Present
Contextual AI / IndulgeOut
Competitive benchmarking of RAG architectures, vector search platforms, and prompt/memory strategies. Built an annotation platform integrating retrieval & generation APIs, improving annotation speed by 50%. Led rapid prototyping of a mobile app including system architecture, APIs, and technical workflows.
Nov 2021 – Apr 2024
AB InBev
Designed a Credit Risk ML model that drove a 12.6% uplift in net revenue. Built forecasting models (ARIMA, Prophet, LSTM) improving packaging accuracy by 12%. Built end-to-end Azure Data Factory pipelines, saving 2 hours of planning time per day. Spearheaded a GenAI LangChain POC for the Logistics domain.
Oct 2020 – Oct 2021
Piramal Glass
Built a predictive model for daily fuel consumption enabling anomaly detection and saving $500k in fuel costs. Automated financial transaction reconciliation using ML, reducing reconciliation time by 30%.
Jul 2016 – Oct 2020
Publicis.Sapient
Performed market mix modelling to estimate effectiveness of marketing and pricing strategies. Developed a BTYD probabilistic model using customer behavioural characteristics, resulting in a 7% increase in campaign effectiveness.
2022 – 2023
LJMU Liverpool
Postgraduate degree specialising in AI, machine learning, and deep learning. Built real-world ML systems as part of the programme.
2020 – 2021
IIIT Bangalore
Intensive programme covering supervised/unsupervised learning, NLP, computer vision, and model deployment.
2012 – 2016
University of Pune
Bachelor of Engineering in Computer Science. Built strong foundations in algorithms, data structures, and software engineering.
04. projects
Built an annotation platform integrating response generation and retrieval APIs for enterprise RAG evaluation. Improved annotation speed by 50% and provided a scalable workflow for model evaluation across LLM providers.
Designed and deployed a Credit Risk ML model at AB InBev to segment customers into risk categories and determine optimal credit limits and payment terms. Drove a 12.6% uplift in net revenue between control and test groups.
Spearheaded a Generative AI POC using LangChain for the Logistics domain at AB InBev. Capable of ingesting data from multiple sources and generating personalised responses to support daily planning.
Built a predictive model for daily industrial fuel consumption at Piramal Glass, enabling efficient anomaly detection and saving the company $500k in fuel costs. Automated transaction reconciliation reducing reconciliation time by 30%.
Implemented time-series forecasting models (ARIMA, Prophet, LSTM) at AB InBev to predict returnable packaging demand, improving data forecast accuracy by 12% and reducing production loss.
05. github
06. blog
A deep dive into building a retrieval-augmented generation system that actually works in production — covering chunking strategies, embedding choices, and latency trade-offs.
A case for building unexpected, delightful Easter eggs into your personal site — and how a hidden CLI command interface can make you memorable to recruiters and engineers alike.
What I learned shipping demand-forecasting models at scale — how ARIMA, Prophet, and LSTM actually compare in production, and why the model is rarely the hard part.
09. contact
Have a project in mind, a question, or just want to say hi? Drop me a message.