A final-year engineering student updates her resume for the tenth time before a placement drive. She adds a keyword here, removes a bullet point there, and still has no real sense of whether it will pass an applicant tracking system (ATS) or catch a recruiter’s eye.
This is the exact problem Himanshi Sharma set out to solve. An engineering student from Ramrao Adik Institute of Technology (RAIT), Navi Mumbai, Sharma developed CareerStack AI, a platform that analyses resumes, evaluates ATS readiness, identifies missing skills and recommends learning pathways based on a user’s profile.
Her aim was to connect these steps so a job seeker could understand their current profile and what they could do next.
Sharma had studied AI/ML at university, but her experience with Samsung Innovation Campus taught her to begin with the problem and the user before choosing the technology, she tells AIM.
Building CareerStack AI
“CareerStack AI was built to address a common challenge faced by students and job seekers to understand how well their skills and experience align with industry requirements and what steps they should take to improve their career opportunities,” Sharma tells AIM.
The platform begins with a user’s resume. It evaluates the resume’s ATS readiness and extracts information about the person’s skills and experience. It then uses that information to identify skill gaps, recommend roles and generate learning pathways.
Sharma said this connected approach was important because career decisions rarely stop at resume preparation.
“Following the design-thinking approach I learned through Samsung Innovation Campus, I focused on understanding the problem from the user’s perspective first,” she says.
On the technical side, CareerStack is built as a full-stack application. Sharma used Next.js and React for the frontend and Python and FastAPI for the backend.
Its AI capabilities use LLMs hosted through Groq, along with natural language processing techniques for skill extraction. The platform also uses real-time job market data for comparisons with industry requirements.
A typical end-to-end workflow covers resume analysis, skill-gap assessment, roadmap generation and chatbot interactions.
Sharma estimates that such a workflow consumes about 6,500 input tokens and 2,300 output tokens, or roughly 8,800 tokens in total.
The platform has not been monetised yet.
Sharma’s current focus is validating the problem, developing the core product and determining whether the recommendations genuinely help users make career decisions. She sees future possibilities in personalised mentorship, recruiter-facing products and subscription-based career guidance.
Making One System Work
Sharma tells AIM that the most difficult part of building CareerStack AI was not any single feature. It was integrating resume analysis, skill extraction, role matching, and roadmap generation into one coherent workflow. Each of these stages required different data-processing steps.
A skill gap identified during resume analysis had to translate meaningfully into a role recommendation. That recommendation then had to connect to a realistic learning pathway.
Personalisation made the problem harder. A useful recommendation requires the system to understand a user’s experience, compare it with industry requirements and turn that comparison into something actionable.
Sharma addresses this by connecting the different stages of the platform. Resume analysis feeds into skill-gap assessment, which then informs role recommendations and learning pathways.
She says this connected structure is what separates CareerStack AI from standalone resume analysers, job boards, or generic chatbots.
“It provides users with a connected view of their current skills, gaps, suitable opportunities, and recommended next steps,” Sharma adds.
Not Her First AI Project
CareerStack was not Sharma’s first AI project.
Sharma says her university coursework gave her the technical foundation for AI/ML. The unknown road for her was to figure out how to apply that knowledge systematically to problems outside the classroom.
“Samsung Innovation Campus strengthened my ability to approach problems systematically through a ‘question-first’ mindset, focusing on understanding real-world challenges before considering data, algorithms, or technology selection,” she says.
The programme introduced her to frameworks such as CRISP-DM, which helped her move through a project from understanding the problem to developing and deploying a solution. “Earlier, I would often think about algorithms first. Through SIC, I learned to start with the problem, understand the user, and then select the most appropriate technology,” Sharma said.
The programme also introduced her to Design Thinking, an approach built around empathy, iteration, and prototyping. Instead of relying purely on technical assumptions, Sharma started by identifying user pain points first. She built prototypes, tested them, gathered feedback, and refined from there.
Before that, she worked on AI-driven analytics systems, including one focused on HR-related spending. The system analysed spending, identified unusual expense patterns and assessed financial risk while producing AI-powered insights and visual reports around cost optimisation.
She is also working on a research project with a college mate called ARROW-RAG, short for Adaptive Retrieval, Reasoning, Observation, and Withdrawal for Reliable Agentic Generation.
The project focuses on improving the reliability of retrieval-augmented generation systems through adaptive retrieval, evidence verification and claim validation.
The work gives Sharma exposure to a different side of AI development. CareerStack is a product built around a user problem, while ARROW-RAG deals with the reliability of AI systems themselves.
Where AI Can Do the Most Good
Sharma sees education as the area where she would like to spend more time.
She also sees e-commerce as an important area for AI in India. There, she points to applications such as product recommendations, demand forecasting, inventory optimisation, customer support, fraud detection and seller analytics.
Education interests her for a different reason. “In education, AI can address the scale problem,” Sharma says. She believes AI can provide personalised learning paths, identify knowledge gaps and offer targeted feedback to students who may not have access to quality mentoring.
For Sharma, the potential impact extends beyond the individual student.
“Education creates a foundational impact; when people gain relevant skills, access better learning opportunities, and make informed decisions, the benefits extend beyond individual success to broader economic growth and workforce development,” she adds.
Sharma came across Samsung Innovation Campus through her college group. After registering, she went through an MCQ-based assessment and was selected for the programme.
Samsung Innovation Campus is Samsung India’s programme for training young people in emerging technologies. Its 2026 edition aims to train 20,000 people across 10 states, in partnership with the Electronics Sector Skills Council of India and the Telecom Sector Skill Council.
Samsung says the programme has trained more than 26,500 people in India since its launch in 2022. The programme targets people aged 18 to 25 and combines technical training with project-based learning and mentorship.
For Sharma, the most useful part of the experience was how it changed the way she approached a problem. This shift can be seen in CareerStack’s design.
The product is still early and has not yet been monetised. For now, Sharma is continuing to test whether the idea works for the people she built it for, while exploring what she can build next.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.