Career intelligence for the Indian workforce · Build for Bharat 2.0
Vriddhi is three independent engines joined by one integration layer. Career Intelligence tells you which careers are worth pursuing in India over the next five years. Skill Gap reads your resume and tells you how far you are from that career and what to learn first. Job Matching & Salary finds live openings in your city, ranks them by how well you fit, and tells you what to ask for.
Indian postings and INR salaries carry the market signal. Global postings give a comparison baseline. O*NET's task statements let us judge AI exposure task by task instead of guessing. Everything is public, has its source, licence and access date recorded, and loads into a zero-setup SQLite store so the demo runs anywhere without a cloud database.
Indian market · primary signal
72,691 verified postings from three Indian sources, including Naukri, covering the 2024–2026 hiring cycle: titles, skills, experience bands, cities and salary ranges.
Details{
"title": "Senior Data Scientist",
"location": "Bengaluru",
"experience": "4-7 years",
"skills": "Python, TensorFlow, SQL,
Spark, A/B testing",
"salary": "₹18-28 LPA",
"posted": "2025-08-15"
}
Compensation
43,374 normalised salary points in INR (lakhs per annum) and USD, across Bengaluru, Hyderabad, Pune, Mumbai and Delhi-NCR. The baseline for every salary range Vriddhi shows.
Details{
"role": "Data Scientist",
"city": "Bengaluru",
"experience_years": 3,
"salary_inr_lpa": 16.5,
"source": "india_tech_salaries"
}
Global baseline
115,000 cleaned LinkedIn postings (public Kaggle datasets). Used to compare India against the global market, never as a substitute for Indian data.
DetailsOccupation science
1,016 occupations broken into 18,838 official task statements. Lets the AI-exposure model reason about each task a job involves, rather than making claims about whole careers.
DetailsLive · used by engine 3
Current openings by role and city from the Adzuna India API, with JSearch as a fallback. Cached for six hours so repeat searches are instant and the demo never depends on Wi-Fi.
DetailsThe user
PDF, Word or plain text, including scanned resumes. Read in memory and never stored.
Detailsjob_postings_india 72,691 postings job_postings_global 115,000 postings salary_benchmarks 43,374 points, INR & USD skill_demand 367,395 regional skill observations occupations · occupation_tasks · occupation_skills · ai_exposure_tasks // views: v_india_vs_global, v_skill_comparison
Scores every career on today's demand, its five-year growth, how AI changes the work, and how confident the evidence is. Then it ranks careers for you and explains each recommendation in plain language, with the numbers behind it.
How much hiring is there today?
Four signals, not just posting counts: volume (log-scaled so huge generic titles don't drown specialist roles), how many different employers are hiring, how recent the hiring is, and how many cities it spans.
DetailsD = 0.40·volume + 0.25·employers + 0.20·velocity + 0.15·geography // each signal scaled to 0–1
How does AI change this job?
Every career is split into its official tasks. Each task is classed as automated, augmented or human-centric, and the role score is the average. No "this job will disappear" claims.
Detailstasks analysed 26 augmented 18 // AI makes the worker faster exposure 0.45 // augmentation sweet spot
What happens over five years?
Growth combines today's demand, AI leverage and hiring momentum. Yearly trajectories run to 2031 for India and globally, with uncertainty bands that widen the further out they go.
DetailsG = clamp(0.70·D + ΔAI + momentum) ΔAI = (0.50 − exposure) × 0.30 band = ±(1 − confidence)·14·√years
Which career is best for me?
Careers are ranked on weights you can change. Each comes with plain-English reasons and a knowledge graph of its tasks, skills and tools, which engine 2 uses as the target to measure you against.
Detailsoutlook Strong Growth demand 0.91 growth 0.76 confidence 0.95 India median ₹18.5 LPA · Bengaluru, Hyderabad, Pune top_skills python, machine_learning, sql, deep_learning, cloud, docker
Sample output
Career scorecard · Data Scientist (sample)
Ranking weights (default, user-adjustable)
Turns a resume into an evidence-backed skill profile, compares it with the career engine 1 recommended, and returns three things: what you already have, what you're close to, and what you're missing. Then it gives a verdict and a learning roadmap in the right order.
Get the profile in
PDF, Word and text. Scanned pages with no text layer are read with OCR. Sections are split (skills, experience, projects, education) and years of experience are worked out from date ranges, without double-counting overlapping jobs. Or just type your skills.
Detailsok text PDF · scanned PDF (OCR) DOCX · TXT · typed skills rejects corrupt · encrypted · oversized // structured error, never a crash stores nothing // resume text is discarded
Find the skills
A 482-skill taxonomy with aliases catches the common skills fast. An optional LLM pass finds niche ones and is checked against the text, so it can't invent skills. Every skill maps to the same ids engine 1 uses.
Detailsin "ML", "M.L.", "machine-learning" out machine_learning in "PostgreSQL" → postgresql, maps_to sql in "go the extra mile" → not the Go language
Mentioned is not expert
Each skill gets a level and a confidence from where it shows up: only listed, used in a project, or used at work. A skill you claim but never used gets flagged for verification instead of being trusted.
Detailspython L4 0.90 work + project sql L3 0.90 work xgboost L2 0.75 project docker L1 0.50 mentioned only experience 7.5 yrs // overlaps counted once
How far am I?
Each required skill is matched, adjacent (you know something close: PyTorch for TensorFlow, Tableau for Power BI) or missing. Adjacent skills earn partial credit. The verdict says whether you're under-skilled, a good fit, or over-qualified and should aim higher.
Detailsskills Python, SQL, pandas, PyTorch target Data Scientist
match 0.52 verdict under-skilled matched python, sql, pandas adjacent tensorflow ← you know pytorch missing machine_learning, statistics
What do I learn first?
Missing skills ordered by importance, with prerequisites pulled in first (statistics before machine learning, Docker before Kubernetes). Time is shown as an estimated range, shorter for skills you're already close to.
Details1 statistics ~25–40 h 2 machine_learning ~50–80 h 3 tensorflow ~12–20 h // adjacent, halved // estimates, not guarantees
Pulls live openings in your city, scores each one against your profile, and ranks them by fit, not by salary alone. For any job, it estimates the market range, your own range, and a sensible number to negotiate towards.
What's open right now?
Adzuna India first, JSearch if it fails, and our own Indian postings if both are down. City names are normalised (Bengaluru = Bangalore, Delhi-NCR = Delhi + Noida + Gurugram) and duplicates across sources are removed. Every job keeps its source, apply link and posting date.
Detailscache (6 h) → Adzuna India → JSearch → Indian postings snapshot // the demo can't break on stage
Which ones fit me?
Each job is scored on skills, experience, education, location, seniority and your preferences, with configurable weights. You see your fit for every job and exactly which skills that job wants that you don't have.
Detailsskill match 92% experience match 67% // 2 yrs vs 3+ asked overall 84% Good match missing here redis
What am I worth?
A range with a confidence score, never false precision. Built from role, city, experience, skills, seniority and how well you match, against the INR salary benchmarks.
Detailsestimated ₹13.0 – 16.0 LPA median ₹14.5 LPA confidence 0.72
What do I ask for?
Compares what the job posts with the market and with you specifically, then suggests a target and a sensible minimum, with the reasons behind them.
Detailsposted ₹12 LPA market ₹13 – 16 LPA you ₹13.5 – 15 LPA ask for ₹14.5 LPA because strong skill match, high-demand specialisation
The engines never import each other's code. They talk through versioned REST contracts, so any one can be replaced or go down without breaking the rest. The integration layer chains them into one journey and adapts their outputs.
One journey
Discover a career, assess yourself against it, find jobs, estimate your value, then improve and repeat. If one engine is unavailable, the others still answer and the gap is shown honestly.
DetailsIntegrate it
Every engine is its own FastAPI service under /api/v1, with Swagger docs generated automatically, so HR platforms and colleges could plug in directly.
Details// engine 1 POST /api/v1/career/analyze POST /api/v1/career/rank POST /api/v1/career/search_by_domain GET /api/v1/career/compare // engine 2 POST /api/v1/skills/analyze_resume POST /api/v1/skills/gap_analysis // engine 3 POST /api/v1/jobs/search
Talk to it
A conversational layer on top of the three engines. The LLM decides which engine to call, and answers only from what they return, citing them.
DetailsTry it
Enter your skills, experience, a target role and a city. The demo walks through the same steps the real system takes, using illustrative sample data.
Knowledge graph
Engine 1 builds a graph for every career: its domain, tasks, skills and tools. Engine 2 uses it to work out prerequisites and adjacent skills. This is a small illustrative slice. Hover over or tab to a node to see what it connects to.