Jul 2022 – Jul 2025 · 3 yrs · Singapore · On-site

• Leveraging a diverse background in technology and innovation, spearheaded the development of Flomoney, a startup dedicated to revolutionizing the legal tech industry. Worked previously with Chope and Crayon to focus on building specialized chatbot solutions, culminating in the creation and launch of LawGPT.
• Chope | Senior Data Scientist
• July 2022 - February 2023
• Played a pivotal role in digitizing menus and enhancing the dining experience through interactive chat capabilities. Successfully led short-term projects focused on menu digitization, laying the foundation for future endeavors in the tech industry.
• Crayon | AI Engineer
• June 2023 - November 2023
• Applied expertise in working with large language models to develop innovative applications for Crayon, an Austria-based company. Collaborated with cross-functional teams to build cutting-edge solutions, laying the groundwork for subsequent projects at Flomoney.
• Microsoft Founder Hub | Selected Startup
• February 2023 - Present
• Secured funding and support from Microsoft Founder Hub to pursue entrepreneurial endeavors. Embarked on the journey to develop FloBot, a versatile chatbot platform, and subsequently transitioned into the legal tech space with the inception of LawGPT.
• LawGPT | Current Project Lead
• November 2023 - Present
• Led the development and launch of LawGPT, a specialized AI model tailored for legal professionals. Currently overseeing the product's review with Singapore-based law firms and spearheading efforts to drive adoption and investment opportunities. Reach out to schedule a demo or explore investment opportunities!

Aug 2021 – Jul 2022 · 11 mos · Singapore

• At Mediacorp, as a Senior Manager, I was responsible for developing a recommendation system for news articles. I implemented a graph-based recommendation system, deploying it to a Kubernetes server for real-time recommendation.
• I also used Deepwalk by Perozzi et. al. to develop recommendation models for sparse graphs, as well as a similarity search similar to Facebook's FAISS for querying similar vectors in real-time.
• I further trained and tested the recommendation models on Databricks and deployed them to a Kubernetes server, creating an end-to-end recommendation pipeline. This included getting training and test data from blobs, spark-based pre-processing, training and evaluation, deploying to production server, and A/B testing.
• To optimize the hyperparameters for real-time incoming data, I implemented a faster version of grid search based on the penetrative grid search model.

Nov 2020 – Jul 2021 · 8 mos · Bengaluru, Karnataka, India · Hybrid

• At Flomoney, we consulted for three companies: Palo-IT, Palpx, and Charles & Keith Group. We were tasked with converting data science knowledge into real algorithms.
• For Palo-IT (Nov, 2020 - Jan, 2021), we built a configurable generic recommendation system using Python, YAML, and Pandas, which could be easily experimented, deployed, and evaluated by a Data Scientist.
• With Palpx (Feb 2021), we built a smart aquaponics setup based on the merger of IoT, chatbot, and aquaponics technology, and published a paper with Subhash Kunnath and Abinash Sahu in PaKDD 2021.
• For Charles & Keith Group (Mar 2021 - Jul 2021), we developed a Django-based dashboard for creating a feedback form and analysis, as well as a consolidation and replenishment model for offline stores. Additionally, we worked on rate of sales modelling for Charles & Keith products, using LightGBM, Bayesian optimisation, and interpretability based on LIME.

May 2018 – Oct 2020 · 2 yrs 5 mos · Hangzhou, Zhejiang, China

• At Aliexpress, I joined the Search and Recommendation team and worked on traffic and growth.
• My work included deep reinforcement learning, actor critic networks, policy gradient algorithms, unsupervised learning algorithms such as VaE and VaDE, multitasking multigated models and adversarial training, evolutionary strategies as a scalable alternative to reinforcement learning, PU learning to label data from positive unlabelled data, PS_smart, an Alibaba-based algorithm to train a reranking model, and a pricing model for Aliexpress based on Porter's five forces model.

Aug 2016 – Apr 2018 · 1 yr 8 mos · Pune Area, India

• At Sears, I worked as a Lead Data Scientist, guiding a team of two Data Scientists and two Data Engineers.
• My team and I worked on machine learning models to create an automated model on member retention, defining store health based on demographic, transactional, store level, and inventory level features, conceptualizing the notion of good versus bad stores, working on convolutional neural nets using torch to determine pregnant shoppers, and using variational autoencoders and Variational Deep Embeddings for clustering customers visiting a particular store.
• We also published a paper titled "Predicting Pregnant Shoppers Based on Purchase History Using Deep Convolutional Neural Networks" in JAIT.

Oct 2015 – Aug 2016 · 10 mos · Bengaluru Area, India

• Solely responsible to build the chat based architecture from scratch. Hired a data scientist from IIT Bombay and guided him to build more algorithms for data science team as well as acting as a bridge between data science and backend team consisting of experienced developers with past experience from leading retail companies.
• I was responsible to write a code for automating chats and enabling virtual agents to give customer a fast, efficient and human-like experience.

Jan 2014 – Sep 2014 · 8 mos · Mumbai Area, India

• Basically worked as a senior software engineer for the Talk to a Teacher initiative by IIT Bombay as India's biggest MOOC. Worked for a period of 9 months and developed a biometric system that could recognize fingerprint of students appearing for tests from remote location using a low end webcam in limited lighting. Due to the research value the work got accepeted in a leading conference in spain.
• I worked to develop a biometric based recognition system from scratch for students enrolled in MOOC conducted by IIT Bombay. The recognition system is unique in the sense that it uses webcams to capture fingerprints compared to conventional mechanical fingerprint sensors. I published a paper titled "Using Webcam to Enhance Fingerprint Recognition" in AMDO, 2014 held in Spain.

Sep 2009 – Mar 2010 · 6 mos · university of lisbon

• I worked as a researcher on NLP,information retrieval in the Grease Project.

Jun 2009 – Sep 2009 · 3 mos · Hyderabad, Telangana, India

• Worked as a researcher on solar direct pumps and developed working models of pumps run by solar energy even at low isolation.

May 2008 – Jul 2008 · 2 mos · Rocquencourt, Ile-de-France, France

• Worked under Prof Gerard Huet and then his PhD and now Prof Benoit Razet. Working in functional programming language such as ocaml to develop a parser for different types of automaton.