profile-pic

Dulal

Innovative and forward-thinking professional with 11+ years of experience, blending expertise in Generative AI (GenAI), Artificial Intelligence and Machine Learning (AI/ML), and Data Engineering with a strong foundation in traditional software application development. Currently advancing cutting-edge research and practical implementation of AI-driven systems while pursuing an M.Tech. in Artificial Intelligence and Machine Learning, effectively balancing academic pursuits with ongoing professional responsibilities. Demonstrated success in designing and implementing scalable, data-driven/business impact-driven applications, embedding AI/ML tools into product solutions. A strategist with a deep understanding of Language Translation, Content Localization, and Speech Technologies. Driving innovation in multilingual and AI-driven applications through active research and work on end-to-end ASR systems, Generative AI (GenAI), and Prompt Engineering. Having good experience of analysing and exploring Large Language Models (LLMs) and evaluating Neural Machine Translation (NMT) models for Indic Languages.
  • Role

    Sr. Software & Meteor Developer

  • Years of Experience

    12 years

  • Professional Portfolio

    View here

Skillsets

  • LoRA
  • Pruning
  • prompt tuning
  • Prometheus
  • Prefix tuning
  • Ppo
  • PostgreSQL
  • Policy gradients
  • nginx
  • NER
  • NEMO
  • LSTM
  • Q-learning
  • Language identification
  • LangChain
  • Indictrans2
  • Indexing
  • Hugging Face
  • Grafana
  • Gradio
  • GitHub Actions
  • gans
  • Flow-based models
  • Flask
  • SQLScript
  • Schema re-architecture
  • Flowgraphs
  • Vector reranking
  • Marian transformer
  • Speech data collection
  • Zero-shot learning
  • whisper
  • Wgan
  • Wer
  • Top-k sampling
  • Streamlit
  • FlashAttention
  • sentiment analysis
  • Sarsa
  • SARIMA
  • SAP HANA Cloud
  • Rlhf
  • Restful APIs
  • react
  • rag
  • Quantization
  • QLoRA
  • Microservices
  • Indicconformer
  • Indicwav2vec
  • Views
  • Triggers
  • Transformers
  • Rnn
  • pandas
  • NumPy
  • Nmt
  • MLOps
  • MLFlow
  • Actor-critic
  • Kubeflow
  • GenAI
  • Cnn
  • ASR
  • Agile
  • TensorFlow
  • Scikit-learn
  • Reinforcement Learning
  • PyTorch
  • Kubernetes
  • Kubernetes
  • contrastive learning
  • few-shot learning
  • federated learning
  • FastAPI
  • FAISS
  • ETL pipelines
  • DVC
  • Dqn
  • DPO
  • distributed ML
  • Distillation
  • Dependency Parsing
  • Docker
  • Cer
  • BM25
  • Beam search
  • Azure ML
  • AWS S3
  • AWS EC2
  • Autoencoders
  • ARIMA
  • Annotation
  • adapters

Professional Summary

12Years
  • Apr, 2020 - Present5 yr 6 months

    Senior Developer

    SAP Labs India
  • Mar, 2016 - Mar, 20204 yr

    Senior Software Engineer

    Tricon Infotech
  • Mar, 2015 - Nov, 2015 8 months

    Software Engineer

    US Tech Solutions
  • Oct, 2013 - Dec, 20141 yr 2 months

    Software Developer

    Magna Infotech
  • Dec, 2014 - Feb, 2015 2 months

    IT Operations Analyst

    Accenture

Applications & Tools Known

  • icon-tool

    PyTorch

  • icon-tool

    Scikit-learn

  • icon-tool

    Hugging Face Transformers

  • icon-tool

    Docker

  • icon-tool

    Kubernetes

  • icon-tool

    SAP HANA

  • icon-tool

    Java

Work History

12Years

Senior Developer

SAP Labs India
Apr, 2020 - Present5 yr 6 months
    Led design and deployment of ASR pipelines for Hindi and English using NVIDIA NeMo Conformer, IndicWav2Vec, and IndicConformer. Curated diverse datasets covering read, spontaneous, and conversational speech. Fine-tuned NeMo models to reduce domain mismatch. Benchmarked CER, WER, latency, real-time factor, GPU/CPU utilization, and scalability. Deployed training workflows on AWS multi-GPU cloud infrastructure. Developed speech data collection platform and re-architected SAP Translation Hub migration with Kubernetes deployment. Evaluated NMTs for Indic languages and improved short-text language detection for enterprise scenarios.

Senior Software Engineer

Tricon Infotech
Mar, 2016 - Mar, 20204 yr

Software Engineer

US Tech Solutions
Mar, 2015 - Nov, 2015 8 months

IT Operations Analyst

Accenture
Dec, 2014 - Feb, 2015 2 months

Software Developer

Magna Infotech
Oct, 2013 - Dec, 20141 yr 2 months

Major Projects

8Projects

ShrutiLekha Indic Speech Data Collection Platform

    Developed a full-stack speech data collection and review tool for Indic ASR. Features include prompt assignment, recording interface, reviewer scoring, and S3 storage integration. Built with React, Flask, PostgreSQL, Docker, AWS EC2/S3, and NGINX.

Translation Hub

    Re-architected SAP Translation Hub migration from Neo HANA Cloud with Kubernetes deployment. Designed schemas, tables, stored procedures, views, indexes, and flowgraphs in HANA Cloud. Built ETL pipelines and migrated legacy services into Java.

Evaluating NMTs for Indic Languages

    Conducted comparative analysis of SAP NMT vs IndicTrans2 for English-Hindi translation. Evaluated translation quality using BLEU, METEOR, TER, ChrF, and COMET. Fine-tuned Indic models and experimented with ensemble approaches.

Investigation on Short Text Language Detection

    Improved language detection for tweet-length SAP text. Benchmarked LangDetect, Ldig, and FastText. Created new short-text datasets and fine-tuned models for accuracy and robustness.

Multi-Agent VQA (VideoMultiAgents Reproduction), Research Work

    Reproduced and extended Stanford's Multi-Agent VQA framework on Kubeflow GPU cluster. Preprocessed NExT-QA dataset and integrated VidOR IDs. Benchmarked single-agent baselines vs multi-agent protocols and optimized with caching and worker scaling.

Conversational AI Systems

    Built and compared Conversational AI pipelines. Evaluated LoRA vs Prefix Tuning and RAG. Implemented hybrid RAG retrieval with cross-encoder reranking and guardrails. Deployed Financial QA chatbot.

MLOps Pipeline California Housing

    Developed end-to-end MLOps workflow for housing price prediction. Implemented data/model versioning and FastAPI deployment. Automated CI/CD and integrated monitoring with Prometheus and Grafana.

Partitioning AI Models for Wireless Latency Prediction

    Explored partitioning strategies for latency prediction in wireless networks. Implemented vertical and horizontal partitioning and proposed a hybrid edge-cloud strategy for scalable 5G deployment.

Education

  • M.Tech. Artificial Intelligence and Machine Learning

    Birla Institute of Technology and Science
  • B.Tech. Computer Science Engineering

    Punjab Technical University (2012)

Certifications

  • Microsoft certified: azure data scientist associate

  • Prompt engineering for chatgpt