Nov 2025 – Jun 2026 · 7 mos · Bengaluru, Karnataka, India · On-site

• Working on building the AI layer for the company - I have designed a lot of AI systems from scratch here, including but not limited to:
• Knowledge Engineering & Semantic Ingestion: Architected a centralized knowledge ingestion and semantic indexing platform for large healthcare corpora with document-specific preprocessing, chunking strategies, and embedding pipelines. Built LLM extraction pipelines that convert long regulatory documents (e.g., clinical standards and policy texts) into structured, taxonomy-tagged rules and knowledge.
• Agentic Workflows & Information Extraction: Developed multi-step agentic pipelines orchestrating LLM reasoning across tasks such as document classification, rule abstraction, and spoken clinical transcript processing. Systems detect entities (e.g., patients), extract tasks from care plans, and generate structured notes for downstream healthcare systems.
• Retrieval-Augmented Reasoning Systems: Built multilingual RAG architectures combining dense and sparse
• retrieval, query reformulation, intent classification, and user-memory integration. Enables grounded question
• answering over healthcare literature with inline citations and bibliography generation.
• LLM Evaluation, Alignment & Data Infrastructure: Designed automated compliance evaluation pipelines that
• score clinical reports against extracted regulatory rules. Improved production conversational agents through prompt orchestration, guardrails, and token optimization (15% reduction), and built tools for voice/chat data collection and prompt A/B testing.

Jun 2024 – Nov 2025 · 1 yr 5 mos

• At Convai, I worked at the intersection of AI engineering, applied research, and infrastructure — building and scaling cutting-edge LLM/VLM applications. I deployed and served large models (OpenAI, Anthropic, Gemini, LLaMA) on custom cloud infrastructure using vLLM, integrated advanced prompting, function/tool calling, and built guardrails to ensure safe and reliable AI systems. Created Prototypes and features for big companies like lockheed martin, Unilever, Linden Labs etc..
• I designed custom data pipelines and fine-tuned large multimodal models (Llama-3-70B and Llama-3.2-90B-vision) to improve output quality. I also developed a novel semantic/BM25 hybrid chunking method that cut latency by 27% while reducing memory usage by 3–4x, and ran extensive RAG experiments to optimize embeddings, chunking, and retrieval strategies.
• Beyond research, I prototyped advanced AI systems — from multi-agent group chats orchestrated by a “director” model, to browser agents that could automate real tasks, avatar-generation pipelines, and real-time chat plugins that enabled contextual QA over live websites.
• On the infrastructure side, I maintained the internal model registry, streamlined model onboarding (including customer-provided models), and used Postgres/SQL analytics to migrate users seamlessly to new model versions. By optimizing cloud pipelines and serving strategies, I reduced costs by up to 10x while scaling to production workloads.
• I also built evaluation frameworks to auto-generate test cases and benchmark prompts with LLM-as-judge evaluators, ensuring that the systems we built were not only powerful but also measurable and reliable.

Feb 2024 – May 2024 · 3 mos

• Developed a probabilistic framework for dynamic personality modeling in LLMs, incorporating Big 5 personality traits through systematic prompt engineering.
• Implemented and evaluated few-shot prompting techniques to optimize character personality consistency, achieving a very human-like personality
• Engineered synthetic data generation pipelines for fine-tuning smaller LLMs (google-gemma-1.1-2b, microsoft phi-3-3.8B), focusing on emotion perception and contextual response generation.
• Developed comprehensive testing frameworks and evaluation metrics to assess AI character performance and behavioral consistency.

Aug 2023 – May 2024 · 9 mos · Roorkee, Uttarakhand, India · On-site

• Worked on building intelligent Deep learning models for predicting state of charge of sodium ion batteries. Built a custom dataset for Indian battery conditions in electric vehicle

May 2023 – May 2024 · 1 yr

May 2022 – May 2023 · 1 yr

• Senior Manager at the social entrepreneurship club, based on a global non profit organization ENACTUS.
• Assisting the project lead in managing a team of 25+ enthusiastic students in developing and refining social ventures. Currently working in project khoj

Apr 2022 – May 2022 · 1 mo

May 2023 – Aug 2023 · 3 mos · Hamilton, Ontario, Canada · On-site

• Conducted optimization research for EV recycling center placement in the Greater Toronto Area, developing multiobjective cost minimization models.
• Implemented non-linear optimization algorithms in Python, incorporating both financial and environmental cost vectors into the decision framework.
• Performed comprehensive literature analysis of recycling supply chain components, synthesizing findings into an optimization framework for sustainable infrastructure planning.

Aug 2022 – Oct 2022 · 2 mos

• Manage the team of developers, testers and designers to successfully deliver the product to our clients.
• Handle client meetings and closely coordinate with all the stakeholders to effectively plan the sprints, establish timelines and tasks.

May 2022 – Jul 2022 · 2 mos · Bengaluru, Karnataka, India

• Worked in data collection and research for a professor in the field of strategic entrepreneurship.
• The paper was based on the effect of human capital signals of the startup founder on the funding raised. Different cross sectional analysis were done, and effects of various institutional strengths and control variables were studied

Sep 2021 – Apr 2022 · 7 mos · Roorkee, Uttarakhand, India

Apr 2021 – Sep 2021 · 5 mos · Roorkee, Uttarakhand, India

Jun 2021 – May 2022 · 11 mos · Roorkee, Uttarakhand, India

• Deputy Cell Secretary of the Videography & Photography Cell, NSS IIT Roorkee, the largest student body responsible for the welfare of society and the Nation.
• ● Assisted in spearheading a combined workforce of 25 students ensuring a good social media presence and welfare work.
• ● Produced digital solutions for the cause of the Women's Empowerment, Healthcare and Historic Movements.