Tejas Arya

Strong match for fast growing startups and product companies hiring through Instahyre for AI, backend, platform, and applied ML roles. Candidate has production experience building scalable AI systems, LLM applications, retrieval augmented generation pipelines, inference systems, and distributed backend services. Demonstrates strong ownership, system design skills, problem solving ability, rapid execution, startup adaptability, and end to end product development experience.

Suitable for roles involving generative AI, machine learning infrastructure, backend engineering, applied research, cloud deployment, and scalable software systems.

Applied Scientist I at Tekion Corp (2024-07 – Present)

• Architected an end-to-end Multi-Agent Conversational Bot using LangGraph to automate inbound SMS for dealerships, handling inventory discovery, vehicle recommendations, and transactional workflows.
• Implemented Tool Calling (Function Calling) to interface with external REST APIs, enabling agents to dynamically query Dealership Timings and execute Appointment Bookings within the conversation.
• Leveraged a Knowledge Graph as the primary state store to model relationships between vehicle inventory and lead context; developed a recommendation engine using Graph Embeddings to suggest relevant alternatives.
• Developed a Vehicle Search React Agent that dynamically routes queries to the appropriate tool — SQL for structured analytical questions (e.g., pricing filters, availability counts), vector embeddings for fuzzy/semantic search, and additional tools (e.g., knowledge graph lookup) based on query intent, enabling natural language vehicle discovery across diverse query types.
• Productionized an LLM-driven Call Intent Classification platform as event-driven microservices using Kafka and Docker, processing 10,000+ transcripts/day with robust backpressure controls.
• Integrated MLflow on Databricks to establish a centralized evaluation framework, tracking model drift, response quality metrics, and API reliability through automated precision/recall dashboards.

Applied Scientist I at Tekion Corp (2024-07 – Present)

CRM Pro

• Built an Email Topic Suggestion & Generation service serving 5k+ emails/day, accounting for 30% of total dealership emails and generating $6M annual revenue.
• Surfaced ranked topics from prior interactions; on selection, generated personalized drafts aligned with dealership tone and compliance.
• Implemented conversation summarization that condenses multi-channel history (calls, emails, chats) into high-signal context passed to the generator.
• Owned production operations: maintained services, debugged incidents, and resolved latency issues in real-time draft creation (profiling I/O, trimming prompt size, adding caching).
• Used Kafka to orchestrate events (trigger summarization, fan-out generation) and decouple producers/consumers with backpressure.
• Used MongoDB to persist versioned conversation summaries and context snapshots per lead.
• Used Redis for low latency caching of prompt, deployment keys.

Associate Software Engineer III at Carelon Global Solutions (2023-08 – 2024-07)

• Built NLP pipelines for healthcare data: applied TrOCR to PDFs/handwritten notes (~95% accuracy) and improved NER with LayoutLMv3 (+20% accuracy).
• Fine-tuned BERT for domain QA (~85% accuracy) and integrated a Vector DB to reduce query latency by ~40%.
• Delivered a document QA system with LangChain + RAG, processing 10k+ medical documents/day.
• Automated OpenCV-based text detection to handle 500+ documents/hour; built Streamlit dashboards to surface insights.

Master of Technology in Artificial Intelligence & Machine Learning – Indian Institute of Technology Hyderabad (2021-08 – 2023-05)

Bachelor of Engineering – Netaji Subhas Institute of Technology, Delhi (2017-08 – 2021-05)