Download PDF (1 page) LaTeX Source

Professional Summary

ML Engineer with 5+ years of experience in production ML systems, LLM development, and applied AI research. Published at NeurIPS, AAAI, and ICWSM with 56+ citations. Track record of deploying ML at scale: 15-30% accuracy improvements, 99.89% pipeline uptime, and systems serving 50K+ daily users. Experienced in LLM fine-tuning, distributed training, and end-to-end MLOps. Strong foundation in responsible AI through published work on misinformation detection and algorithmic fairness.

Education

University of California Los Angeles (UCLA) — Los Angeles, CA

Master of Science in Computer Science | GPA: 3.81/4.0 | Sep 2021 - Jun 2023

Indian Institute of Technology (IIT) Kharagpur — Kharagpur, India

Bachelor of Technology in Computer Science & Engineering | GPA: 9.04/10.0 | Jul 2016 - May 2020

Work Experience

Stovell AI Systems — San Francisco, CA

Machine Learning Engineer | Jan 2024 - Present

  • Architect end-to-end ML pipelines for model training, deployment, and monitoring, tracking both model accuracy and business ROI impact across multiple customers
  • Design and train custom architectures (Transformers, LSTMs, CNNs, hybrids) for fuel price and volume demand forecasting, achieving 15-30% MAE reduction
  • Develop causal inference models for price-volume elasticity prediction; create evaluation frameworks measuring model performance and downstream business impact
  • Build production ETL pipelines for data ingestion, processing, and model inference; maintain 99.89% uptime SLA across all customer deployments
  • Implement comprehensive monitoring with DataDog and Weights & Biases for data quality validation, model drift detection, and accuracy tracking across geographies
  • Conduct systematic model evaluation including robustness testing, edge case analysis, and failure mode identification before production deployment

Amazon Web Services (AWS) — San Francisco, CA

Applied Scientist Intern | Jun 2022 - Sep 2022

  • Benchmarked large language models (ProtBERT, ProtT5, ProtGPT) for Drug-Target Interaction (DTI) prediction; integrated diffusion-based molecular docking
  • Implemented data-parallel distributed training on P4d.24xlarge instances (8x A100 GPUs), achieving 2x memory efficiency through gradient checkpointing and mixed precision
  • Improved baseline performance by 20% and 12% on two benchmark datasets using protein language model embeddings; results presented at Amazon ML Conference
  • Developed systematic evaluation pipeline for model comparison including accuracy, inference latency, and computational cost tradeoffs

Goldman Sachs — Bangalore, India

Software Engineer | Aug 2020 - Sep 2021

  • Led full-stack development integrating two internal bug tracking systems with migration to AWS, architecting scalable RESTful APIs serving 50,000+ daily users
  • Built backend services with Java/Spring Boot and frontend with TypeScript/Angular; managed complete SDLC from requirements to production deployment
  • Designed database schemas and optimized query performance for high-throughput issue tracking workflows

Kanini Software Solutions — Los Angeles, CA

DevOps Engineer | Aug 2023 - Jan 2024

  • Automated infrastructure provisioning for healthcare platform using Terraform, improving deployment reliability with AWS S3, Lambda, and RDS

Accenture Technology Labs — Bangalore, India

Research Intern | May 2019 - Aug 2019

  • Developed GCN-based stock prediction model integrating news articles and knowledge graphs, improving MSE by 5% over baseline methods

University of California Los Angeles — Los Angeles, CA

Graduate Teaching Associate | Sep 2021 - Jun 2023

  • Led discussions and office hours for 500+ undergraduates in CS32 (Data Structures) and Chemistry 20A, totaling 500+ teaching hours across 5 quarters

Publications

PGraphDTA: Improving Drug-Target Interaction Prediction using Protein Language Models and Contact Maps Rakesh Bal et al. | NeurIPS 2023 Workshop on AI for Science

  • Developed novel architecture combining protein LLM embeddings with structural contact maps for improved DTI prediction

Two-Sided Fairness in Non-Personalised Recommendations Rakesh Bal et al. | AAAI 2021 Student Abstract

  • Proposed fairness-aware recommendation framework balancing user and item-side equity

Analysing the Extent of Misinformation in Cancer Related Tweets Rakesh Bal et al. | ICWSM 2020

  • Created dataset and attention-based BiLSTM-CRF model for detecting health misinformation; contributed to responsible AI by identifying harmful content patterns

Selected Projects

DeepSub: Fine-tuning LLMs for Multi-Source Subtitle Translation — GitHub

  • Fine-tuned Gemma-7B and LLaMA-7B using QLoRA (4-bit quantization) for multilingual subtitle translation synthesizing context from 2 source languages
  • Implemented triplet translation pipeline; evaluated on 6 Indic and 3 European languages with BLEU score tracking via Weights & Biases

CLIP for Visual Question Answering (VQA) — GitHub

  • Integrated OpenAI CLIP with VQA architectures (MCAN, Pythia) achieving 2% improvement in both zero-shot and fine-tuned settings on VQA2.0
  • Added Language-Driven Semantic Segmentation (LSeg) for improved counting questions

Skills

Category Technologies
LLM/GenAI Transformers, LLM Fine-tuning (QLoRA/LoRA/PEFT), vLLM, LangGraph, HuggingFace, Prompt Engineering, Model Evaluation
ML Frameworks PyTorch, TensorFlow, Keras, scikit-learn, XGBoost
MLOps Docker, Kubernetes, KubeFlow, MLflow, Weights & Biases, Distributed Training, Model Serving
Cloud/Infra AWS (SageMaker, EC2, S3, Lambda), GCP, Azure, Terraform
Data Snowflake, DBT, ETL Pipelines, DataDog, SQL
Languages Python, Java, C++, Go, SQL, TypeScript

Achievements

  • IIT-JEE Advanced: All India Rank 265 out of 200,000 candidates
  • IIT-JEE Mains: All India Rank 187 out of 1,300,000 candidates
  • KVPY Fellowship: All India Rank 191 (prestigious national science scholarship)