May 2023 – Sep 2024 · 1 yr 4 mos · Toronto, Ontario, Canada · On-site

• 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗮𝗻𝗱 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗟𝗟𝗠 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀
• Hosted the Shakudo platform, including Triton, Airflow, MLflow, Ray, Airbyte, and VectorDBs, enabling
• MLOps for 15+ clients across cloud and on-premises environments with production monitoring tools.
• Resolved high-urgency issues (service mesh, certificates, storage) for client clusters, ensuring minimal downtime and robust service continuity. Provided solution engineering support, enhancing client satisfaction.
• Developed software solutions using LLMs for knowledge extraction and question answering on web and
• internal knowledge bases, boosting the productivity of the finance and real estate teams.
• Enabled Sales and Marketing with 20+ ML demos and automated MQL Analytics, securing 6 new clients
• through data engineering, NLP, and vision use cases, plus blogs on vector databases and LLMs.

May 2022 – Oct 2022 · 5 mos · Montreal, Quebec, Canada · Hybrid

• 𝐎𝐩𝐭𝐢𝐦𝐢𝐬𝐢𝐧𝐠 𝐍𝐞𝐱𝐭 𝐔𝐭𝐭𝐞𝐫𝐚𝐧𝐜𝐞 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 & 𝐒𝐩𝐞𝐥𝐥 𝐂𝐡𝐞𝐜𝐤𝐢𝐧𝐠
• Optimized Next Turn Prediction for a Conversational Agent AI App. Tweaked the Transformer models by reducing embedding size and ensemble approaches, resulting in 1% EM improvement with 2X latency reduction.
• Developed code for FastBert, a single-model ensemble in Tensorflow
• Implemented Spell Checker for the AI application using Viterbi Algorithm with 91% F1, improving baseline by 6%. Streamlined decisions on optimally using Bing Spell Checker API by building a prototype
• Learnt excellent practices of research on experimentation, documentation and presentation skills
• Participated and presented in the reading groups where I could think critically and ask questions about others’ Work and the latest research in NLP.

Jul 2020 – Jul 2021 · 1 yr · Pune, Maharashtra, India

• 𝐏𝐫𝐨𝐯𝐢𝐝𝐞𝐫 𝐍𝐞𝐭𝐰𝐨𝐫𝐤 𝐀𝐧𝐚𝐥𝐲𝐬𝐢𝐬
• Developed Provider Network Analysis App to detect anomalies in the health provider database holding hundreds of thousands of provider contracts and tabular records using OCR, BERT-NER and Pattern mining.
• Extracted information from searchable and scanned contracts building a pipeline with Tesseract OCR, antiword, PDFMiner and BERT-NER.
• Reduced latency 10X using distributed computing with PySpark.
• Deployed the App on-prem, citing around 40 critical mismatches of provider network and 1000s of minor ones.
• Experimented with libraries like Fonduer for knowledge base construction and models like CRNNs for OCR from poor-quality documents
• Comorbidities Progression Prediction
• Developed ML algorithms for Predicting the Progression of Comorbidities in patients.
• Extracted and engineered features from a raw and vast patient database.
• Implemented an LSTM model for time series classification.

May 2019 – Jul 2019 · 2 mos · Bangalore Urban, Karnataka, India

• 𝐕𝐨𝐢𝐜𝐞 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 - 𝐀𝐧𝐚𝐥𝐲𝐬𝐢𝐬 - 𝐀𝐧𝐬𝐰𝐞𝐫𝐢𝐧𝐠
• Explored NLP techniques to Analyse a conversation, Build context and Answer the questions
• Framed as a question-answering task and finetuned BERT on the SQuAD dataset and created a WebApp to demo the results.

May 2018 – Jul 2018 · 2 mos · Hyderabad, Telangana, India

• 𝐓𝐨𝐨𝐥 𝐁𝐨𝐱 𝐟𝐨𝐫 𝐭𝐫𝐚𝐜𝐤𝐢𝐧𝐠 𝐢𝐧𝐯𝐞𝐬𝐭𝐦𝐞𝐧𝐭 𝐝𝐞𝐜𝐢𝐬𝐢𝐨𝐧𝐬
• Developed a Web Application on Apache server using Django for maintenance of Fixed Income, helping analysts to add, maintain and track decisions
• Created a registration work-flow allowing a user to log in and register with admin approval and developed an automatic HTML mail for any update of decision which is sent to analysts mapped to that strategy
• Calculated and displayed Analytics like alpha capture and batting average for a selected date range for strategies using template tags. Used manual decorators to moderate the edit rights on decisions.