.png){kind=logo}
Mihika Prasad Gaonkar
Entry-Level Software Engineer specializing in ML and backend systems
Connect with Mihika
Mihika already has a relationship with Reval, so a warm intro from us gets a much better response than cold outreach.
Recommended
Already have an account?
About
Built and deployed a production LLM-based real-time stance detection system for social media, fine-tuning LLaMA 3.1 on A100s with DeepSpeed ZeRO/FSDP and iteratively refining data to handle sarcasm and context-dependent meaning. Also has Kubernetes operations experience (Kafka/Logstash/Elasticsearch observability pipeline) and delivered an OCR automation project during a Worley India internship that saved 20+ hours/week for on-site energy safety stakeholders.
Experience
Education
Key Strengths
- Deployed LLM-powered real-time stance detection system end-to-end (experimentation to production)
- Efficient large-model fine-tuning under memory constraints using DeepSpeed ZeRO + FSDP on A100s
- Systematic misclassification analysis and iterative dataset refinement for sarcasm/implicit sentiment/coded language
- Improved context handling via contrastive pairs where labels depend on prior messages
- Production reliability mindset: metrics-driven evaluation (accuracy/latency/cost/tool-call correctness) with prompt/tool unit tests and production-traffic replay
- Operational monitoring for agents/workflows using structured logging, tracing, and confidence scoring to detect drift/hallucinations/tool misuse
- Built and operated Kubernetes-based distributed logging/observability stack (Kafka/Logstash/Elasticsearch)
- Delivered stakeholder-driven OCR automation reducing manual review by 20+ hours per week
Discover more candidates like Mihika
Search across thousands of pre-screened, high-quality, high-intent candidates on Reval.
Search Talent