Looking for a job

adarsh jha

@adarshjha3

Message

Software Engineer at TBO.com who shipped an AI email service cutting manual review by 60%.

India

Message

At TBO.com, I developed and shipped a distributed microservice that processes customer emails and meeting notes through an LLM API pipeline, cutting manual review by 60%. I also built data ingestion pipelines using Apache Spark, AWS Glue, and Kafka.

At IIIT-Delhi’s Research to Commercialization program, I refactored a Transformer question-answering model, reducing its size by 75% while maintaining 97% accuracy. I also deployed an OCR-backed FastAPI service that handled 100+ concurrent requests.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer – Production Engineering

Current

TBO.com

Jan 2026 - Present (9 months)

Developed and shipped a distributed microservice handling 500+ daily customer emails and 100,000+ meeting notes via an AWS Lambda worker, cutting manual review by 60%. Constructed ETL pipelines using Apache Spark, AWS Glue, and Kafka, and integrated an AI agent with MCP tool-calling into the platform's gateway layer.

AWS Lambda SMTP Outlook Webhooks LLM APIs Apache Spark AWS Glue Kafka Debezium Apache Hudi DynamoDB Parquet S3 MCP Cloudwatch Microservices ETL Regression Testing

ML Engineer Intern

IIIT-Delhi – Research to Commercialization (R2C)

May 2025 - Jul 2025 (2 months)

Refactored a Transformer QA model with LoRA and 8-bit quantization, reducing model size by 75% and doubling inference throughput while maintaining 97% accuracy. Deployed an OCR-backed FastAPI REST service for document parsing, improving extraction accuracy by 23%.

LoRAs 8 Bit Quantization Transformers fastAPI OCR REST APIs Document Parsing Model Optimization

Independent Research Project – Ontology Path Extraction

Oracle x IIIT-Delhi

Jan 2024 - May 2024 (4 months)

Fine-tuned R-BERT with LoRA, span-based embeddings, and negative sampling, outperforming LLaMA-3, Mistral, and GPT-3 by 6 F1 points on biomedical relation extraction benchmarks. Extracted and validated relationships across 25,000+ biomedical entities, cutting ontology path false positives by 20%.

R BERT LoRAs Span Based Embeddings Negative Sampling Relation Extraction Ontology SQL

Education

Degrees, certifications, and relevant coursework

Indraprastha Institute of Information Technology Delhi

B.Tech, Computer Science and Artificial Intelligence

2022 - 2026

Grade: 10/10 in Operating Systems, Computer Networks, Digital Circuits

Pursuing B.Tech in Computer Science and Artificial Intelligence with coursework in distributed systems, data structures, algorithms, computer networks, operating systems, DBMS, and machine learning.