Nov 2022 – Jul 2025 · 2 yrs 8 mos · Bengaluru, Karnataka, India · Hybrid

• Project Category: Batch Processing Framework
• Created a batch processing app for asynchronous job execution. Users submit jobs via a REST API and receive a unique ID for tracking. The app uses Kafka for event streaming and maintains statuses in PostgreSQL. ARGO Workflow orchestrates execution, while Docker containers enhance scalability. This architecture ensures efficient processing with minimal wait times.
• Tech Stack: REST API, Kafka, PostgreSQL, ARGO Workflow, Docker
• Project Category: HR-Chatbot
• Developed an HR chatbot using RAG and Langchain, integrating the Qwen Chat Model for quality inference. Utilized FAISS for similarity search along with VllmOpenAI for inferencing and a Parent Document Retriever for precise retrieval. Implemented Nemo Guardrails for ethical responses and performed analytics with Langfuse. Designed an intuitive interface with Chainlit, deployed on Kubernetes, and packaged as a Docker image, reducing HR workload.
• Tech Stack: Langchain, LLM, Langfuse, Nemo Guardrails, Chainlit, Kubernetes, Docker
• Project Category: Speech-to-Text
• Developed a real-time Speech-to-Text app using GRPC and Whisper for transcription, deployed on Google Cloud Platform with MLServer. The containerized app supports five regional languages to transcribe doctor-patient conversations within all Reliance Foundation Hospitals, Implemented session management for user-specific transcriptions.
• Tech Stack: GRPC, Whisper, GCP, MLServer, Docker, Kubernetes
• Project Category: Speaker Verification
• Built an application to detect potential employee moonlighting in Reliance Call Centers. Leveraged voice samples to create audio signatures in MongoDB, with GRPC services facilitating batch processing. Used Speaker Diarization for speaker identification, generating and comparing embeddings for detection. The app is containerized with Docker and deployed on Kubernetes for scalability.
• Tech Stack: MLFlow, MLServer, GRPC, Azure Blobstore, MongoDB, Docker, Kubernetes

Oct 2020 – Nov 2022 · 2 yrs 1 mo

• Project Category – KYC Automation
• Built Application for identifying type of merchant KYC document, extract data and apply business rules to automate KYC process for merchants, vendors and card holders. Involved in solution design, implementation and supporting new requirements.
• Implemented REST API’s on Spring Boot which accepted KYC Document and called ML functions to perform OCR and Regex Rules. Used Netflix Conductor for workflow automation for Document Identification, Classification, Business Rules and Report Generation.
• Tech Stack Used-> Spring Boot, REST, Python, Machine Learning, NLTK, Microservices, Postgres and Netflix Conductor.

Jul 2019 – Oct 2020 · 1 yr 3 mos

• Project Category – Email classification for auto-ticket generation
• Used Netflix Conductor to automate workflow for reading and parsing emails from a shared mailbox and determining type and context of emails. Workflow includes creating a ticket with appropriate details and automatically sending emails to customers. Currently around 100 shared mailboxes are in production.
• Spring Boot API’s were used to call ML functions, Postgres DB was used to store email metadata. O365 APIs were used for reading and sending emails. Used NLP functions to determine context of email.
• Tech Stack Used -> Spring Boot, REST, Python, Microservices, Postgres and Netflix Conductor.

Jan 2019 – Jun 2019 · 5 mos

• Project Category – Duplicate Invoice Detection
• Made an application to automate detection of duplicate invoices. Invoices are read via Spring Boot Rest API and passed to ML functions to perform OCR and regex rules on invoices. Relevant information is extracted and compared with extracted information of previous documents to flag potentially duplicate invoices.
• Made a portal using React JS for manual comparison of flagged invoices, cancel/continue the payment process by marking them as duplicate/non-duplicate invoices and to send communication to invoice owner. This capability saved around $10m for the US market.
• Tech Stack Used: Spring Boot, Python, Machine Learning, React JS, Postgres DB

May 2018 – Jun 2018 · 1 mo · Bengaluru Area, India

• As part of Artificial Intelligence Team, developed software that classifies a human heart as normal or abnormal and predicts any kind of abnormalities in Heart that is present or very much likely to occur in near future by reading raw ECG scan data as input.

Jan 2018 – Mar 2018 · 2 mos · Banglore

• Android App Development.
• Made - Domestic Help App.
• It is an app that connects people who need household jobs to employer directly.
• All you have to do is write your requirement and wait for the results to match your needs.
• This app is strictly for household jobs and works.

May 2016 – Jun 2016 · 1 mo · Bengaluru Area, India

• Quality Assurance -
• Made tools to enhance the performance of the Hike Messenger App.