v
vasant112

Vasant M

@vasant112

Software Engineer

India
Inglese, Hindi
Alcune informazioni sono riportate in lingua inglese.
Chi sono
I am Software Engineer with experience in building serverless pipelines, AI automation agents, and full-stack applications. I specialize in Python, TypeScript, and cloud architectures like AWS and GCP. I have a strong background in AI/ML, having developed reinforcement-learning crawlers and agentic security gateways.... Continua a leggere

Competenze

v
vasant112
Vasant M
offline • 

Consulta i miei servizi

Applicazioni web full stack
I will build custom ai powered software and saas products solutions

Esperienza lavorativa

QuarterMill

AI Engineer Intern

QuarterMill • Full time

May 2026 - Jun 2026 • 1 mo

- Developed Python automation agents using Celery and the Anthropic SDK to assess document quality and validate archival references, incorporating Unicode normalization for non-Latin scripts - Engineered a multi-stage verification pipeline using Pydantic, Instructor, and Langfuse to validate structured LLM outputs against source documents, making automated document reviews auditable

Tinkering

Software Engineer Intern

Tinkering • Full time

Feb 2026 - May 2026 • 3 mos

- Automated an end-to-end business workflow spanning polling, analysis, action, dispute, and settlement by building a Node.js service with ethers.js transaction signing, eliminating manual intervention across the lifecycle - Built dispute-resolution workflows integrating Dialectica and Tavily APIs with an LLM synthesis layer and weighted confidence scoring - Implemented a SQLite-backed audit log and indexing layer to capture state transitions from polling through resolution, enabling end-to-end workflow traceability

MITS:_Madhav Institute of Technology & Science, Gwalior

Undergraduate Researcher

MITS: Madhav Institute of Technology & Science, Gwalior • Part time

Jan 2026 - Apr 2026 • 3 mos

- Designed a reinforcement-learning web crawler in PyTorch using Q-Learning, LinUCB, and GraphSAGE to adaptively prioritize pages for crawling - Evaluated the crawler against 7 baselines using fixed-seed experiments; the strongest configuration retrieved 100% of relevant pages in 2.56 seconds, achieving a 9.25× speedup over random crawling