Expertise lived everywhere and could be reached nowhere.
It sat in profiles, project histories, and publications, but it could only be searched by keyword and by who you happened to know.
The right person for a requirement was often invisible, because the words on their profile didn't match the words someone typed into a search box. A keyword system can only return what it was told to look for. It can't reason about capability it wasn't given the exact phrase for.
§ 02 · What was built
02 / 04
A platform that matches people across several kinds of evidence at once.
It scores candidates against a structured competency framework, and runs multi-tenant so multiple client datasets stay fully isolated behind one system.
Semantic talent search · tenant viewsub-second over live dataset
Hero Query
“Senior lawyer with cross-border M&A experience, SEBI and FDI regulatory work, and inbound acquisition mandates for multinational clients”
Search mode: Semantic · Client: Meridian Legal · Industry: Law
Ranked Results (Mock)
194%
Arjun Mehta
Partner · Mumbai
Practice areas
Corporate & M&A, Capital Markets, Regulatory
Evidence behind match
Led four inbound acquisition mandates for multinational clients (2021-2025). Published on SEBI disclosure requirements for cross-border M&A and FDI route selection.
289%
Priya Nair
Senior Counsel · Delhi
386%
Rohan Kapoor
Managing Associate · Bangalore
§ 03 · The detail that made it work
03 / 04
A person was never reduced to a single search vector.
The platform held separate semantic representations for their experience, their expertise, and their published thinking, and combined them only at ranking time.
That separation is what let the system reason properly. A query about hands-on delivery work and a query about thought leadership could surface different people from the same dataset, because the platform weighed each kind of evidence independently before merging them into one ranked result.
Most talent search collapses a person into one blob of text, and loses exactly this distinction.
16 years advising issuers on IPOs, QIPs, rights issues...
Capital Markets, Securities Law, Corporate Advisory...
…
Matters
IPO of leading Indian fintech platform · ₹3,800 Cr
QIP for diversified infrastructure conglomerate · ₹2,100 Cr
Rights issue for renewable energy developer · ₹950 Cr
…
Publications
SEBI’s Evolving Framework for REITs and InvITs · 2025
Pricing and Allocation in Indian IPOs · 2024
Post-Listing Compliance for First-Time Issuers · 2024
…
Scoring layer
profile0.50 × .84
matters0.30 × .91
pubs0.20 × .88
composite87.4%
Query
“Senior capital markets lawyer with IPO, QIP, and SEBI listing experience”
Fig 01 · Each evidence layer is scored against the query independently, then combined with configurable weights into one composite match.
§ 04 · What it enabled
04 / 04
What the platform made possible
01
Search by meaning, not keyword
A natural-language requirement matches the right people even when their profiles never use the same words.
02
Track record, not profile wording
Project history and publications are indexed separately from profile text, so people surface on what they have done, not how their bio is written.
03
Explainable ranking, not a black box
Every result shows why it ranked: profile fit, strongest matching project, and reinforcing publication. Reviewers see evidence, not just a score.
04
Sub-second search over a live dataset
Warmup pre-builds in-memory indices on selection, cutting cold-start search from seconds to sub-second across the working set.
05
Structured competency scoring
People are scored against an explicit, firm-defined capability framework rather than ad-hoc tags, keeping assessment consistent.
06
Assessment, not only discovery
Beyond finding the right person, the platform scores capability dimensions and indicative market-value benchmarks, not just a match.
07
Profiles that stay current
Optional web enrichment supplements held data with recent public activity where permitted, so rankings reflect recent work, not stale files.
08
One architecture, many datasets
Multi-tenant isolation lets the same platform serve separate client environments without data crossing between them.
The same approach, behind your perimeter.
This platform was delivered, owned, and extended by the client. We build the same way for UK law firms: sourced, governed, and embedded behind a partner.
